The rigid registration and fusion of pre-operative 3D volumetric data with intra-operative 2D X-Ray images is an important task for many clinical applications, for instance in non-invasive cancer treatment using radiation therapy. One class of 2D/3D registration algorithms solely relies on image intensities. These methods address the registration problem by computing digitally reconstructed radiographs (DRRs) which are compared to real X-Ray images by means of similarity measures. The main contributions are the development of robust and fast optimization techniques and the utilization of modern graphics processing units to accelerate the DRR computation.

The following images show the overlay of the 3D data for initial misalignment (left) and after registration (right).