Planning and Registration Techniques for Image-Guided Robotic Spinal Surgery

Abstract

This work presents the planning and guidance systems of a novel cooperative robot designed for spinal surgery. The latter has been developed for transpedicular fixation, a particular intervention consisting in the immobilisation of two or more vertebrae by means of screws and metal bars. Its conventional clinical protocols have high levels of invasiveness, increased radiation exposure, difficult visualisation of the surgical field and considerable probabilities of screw misplacement. All these problems could be solved by means of a surgical robotic assistant, which would work cooperatively with the surgeon providing stable and accurate instrument placement. A comprehensive review of the state-of-the-art in surgical robotics for spinal interventions is given in this thesis, focusing on the research done in the last ten years. This study shows that spinal surgical robotics is still in an early stage of development and faces considerable challenges, in particular reaching the theoretical accuracy level demanded by clinical studies. A novel surgical planning application is also presented in this thesis along with a new software library for 2D-3D registration. The latter is a key problem of the proposed robotic system, as it permits accurate localisation of the patient by matching his pre-operative (3D) data and intra-operative (2D) x-ray images. Despite that 2D-3D registration has been studied for many years, researchers were faced with a lack of appropriate software, which was effectively solved with the library presented in this work. 2D-3D registration involves the additional sub-problems of calibration and distortion correction of x-ray imaging devices. These were addressed in this thesis proposing a novel algorithm able to solve both, also providing precise reconstruction of 3D points from their 2D projections. The proposed method is simultaneously accurate, robust and fast, fitting well into the requirements of surgical applications. Finally, integration of the robotic system's components was studied in this thesis, reporting the first experimental results on phantoms. Despite that full integration with the presented 2D-3D registration methods has not been carried out, the robotic system was found to be accurate enough for surgical practice. Full integration and tests with cadavers or animals should be the subject of future works.