Eloy Roura1, Torben Schneider2, Pankaj Daga3, Marc Modat3, Nils Muhlert2, Jordi Freixenet1, Declan Chard2, Sbastien Ourselin3, Xavier Llad4, Claudia A M Wheeler-Kingshott2
1Computer Vision and Robotics Group, Department of Computer Architecture, , University of Girona, Girona, Spain; 2NMR Research unit. Department of Neuroinflammation, UCL Institute of Neurology, University College London, London, United Kingdom; 3Centre for Medical Image Computing, Department of Medical Physics and Bioengineering, University College London, London, United Kingdom; 4Computer Vision and Robotics Group, Department of Computer Architecture,, University of Girona, Girona, Spain
In this work, we tested the performance of 4D multi-channel registration of T1-weighted (T1w) scans and diffusion tensor (DT) data to standard space in Multiple Sclerosis and healthy subjects. The Montreal Neurological Institute (MNI) template was chosen as target. We compared results with single-modality registration (3D) to their corresponding template. For healthy subjects the 4D multi-channel approach performed as well as the 3D approach when co-registering T1w and DT images to MNI atlas space independently. For patients the multi-channel approach improved the co-registration in the boundaries of the ventricles and also in the cortex compared to the 3D approach.