Fluoroscopy-based Rigid Registration for Minimally Invasive Surgery

		Registration of 3D volumetric data to 2D fluoroscopic projections, based on our novel intensity-based registration algorithm.

luoroscopy-based 2D-3D registration of volumetric data with a series of calibrated and undistorted projection images has shown great potential in CT-based navigation because it obviates the invasive procedure of the conventional registration methods, which match the actual location of implanted fiducial markers, anatomical landmarks, or a cloud of points on the surface of the anatomy to the corresponding points in the pre-operative model. Paired-point matching has been successfully adapted for finding initial transformation for the later fine-scale image-based registration. In order to achieve shorter running time, and fewer number of projection images, an intensity-based, multi-resolution 2D-3D registration algorithm has been developed. A novel least-squares normalised pattern intensity (LSNPI) similarity measure was proposed to improve the accuracy and robustness. Our experiments on phantom as well as cadaveric spine show that it is feasible for the newly developed algorithm to achieve sub-millimetre accuracy in a realistic setup in less than two minutes.