Registration of 3D diffusion weighted images using maximisation of mutual information

Diffusion tensor imaging (DTI) characterises anisotropic diffusion of water in the human body. For accurate estimation of tensor parameters it is essential that the diffusion weighted images are spatially aligned. In this study, we demonstrate the use of mutual information to register 3D diffusion weighted images. Specifically, diffusion weighted as opposed to non-diffusion-weighted images are used as the reference for registration and a rigid transformation is used to correct the distortions caused in the images because of patient movement. The results demonstrate that the proposed registration scheme can consistently and effectively characterise the motion induced distortion by minimising the mis-registration of images. Potential applications for this method are discussed.