Martijn John van der Bom1, Josien Petronella Pluim1, Matthew Gounis2, Lambertus Wilbert Bartels1
1Image Sciences Institute, University Medical Center Utrecht, Utrecht, Netherlands; 2Radiology, University of Massachusetts Medical School, Worcester, MA, USA
MRI can provide 3D insight and soft tissue contrast for real-time intra-procedural navigation. However, MRI to X-ray registration is technically challenging due to insufficient similar landmarks caused by differences in physical imaging principles. We have investigated a novel technique for improved 3D MR to 2D X-ray registration. A simulated CT is generated from multispectral MRI data using a k-nearest neighbor classification method. This simulated CT than provides sufficient landmarks for MRI to X-ray registration. Classification was based on MR sequences commonly used in clinical practice. Validation in ex vivo experiments shows that this method is accurate and reliable.