Two Convolutional Neural Networks for the rigid and affine registration of two-dimensional CT-MRI images of the human brain

Image registration is the process of matching the coordinate systems of two or more images. Medical image registration has been used in a variety of applications such as segmentation, motion tracking, etc. Recently, the use of deep neural networks has been demonstrated as a useful approach to registration problems. In this article, we propose two separate novel Convolutional Neural Network (CNN) architectures for multi-modal rigid and affine registration of the CT-MRI images of the brain. A dataset consisting of CT-MRI images of 37 subjects was used for training and evaluation of the networks. For both networks, the proposed models achieved a high mutual information value between predicted CT images and their corresponding MRIs and a mean dice score of 0.984 for rigid registration.