Röntgenbasierte Bildgebung mittels eines roboterisierten C-Bogen-Systems

Christoph Bodensteiner

Abstract

This dissertation proposes methods for the improvement of X-ray and CT-imaging using a novel robotized C-Arm. The discussed methods are purely image-based in order to allow for enhanced imaging without having to improve the hardware parameters of the C-arm (e.g. positioning accuracy, stiffness or detector properties). The fundamental element of the developed methods is the combination of reconstruction and registration techniques. This element has already been successfully employed in different domains of the imageand signal-processing fields. This thesis can be divided into two main parts. The first part gives an overview of the C-arm system used in the course of this work. The construction and kinematics of the system are described, as well as a number of applications that profit from robotized X-ray imaging. Some of these applications require that the geometric distortion of the imaging devices is determined. For the completeness of the work, the implemented method for the correction of the non-linear distortions is detailed. Subsequently, established methods for 3D reconstruction from cone-beam projection are presented and state-of-the-art rigid and non-rigid registration techniques are described. The optimization techniques employed in the implementation of the proposed methods are discussed in detail. The first main part concludes by introducing a new method for image-based correction of imaging systems extrinsic parameters. The second part of this dissertation deals with two applications of combined reconstruction and registration. The first application, super-resolution, involves the acquisition of multiple, slightly displaced images of the same scene in order to compute an X-ray image with increased resolution. Super-resolution has received little attention so far in the field of imaging with mobile C-arms due to the necessity of manual positioning of the C-armdetector. Robotized C-arms offer an elegant approach for the acquisition of the required images, as is illustrated in this thesis, making super-resolution for X-ray images practicable. The second application consists of a motion compensation method for radio-surgery which is based on an intensity-based non-rigid 2D/3D-registration of a patient specific organ deformation model with intraoperative projection data. The deformation model is constructed from a preoperative 4D-CT of the patient using Active Shape Models. Via the incorporation of the statistical knowledge in the combined registration-reconstruction algorithm, the number of needed projections can be greatly reduced. The proposed applications were prototypically implemented and were tested using synthetically generated images as well as images acquired with a robotized C-arm and real patient data. In all applications the 3D-reconstruction errors were significantly reduced when employing the method proposed for data-based correction of the extrinsic parameters of the imaging device. As a result, the quality of images was substantially increased. In the images reconstructed with super-resolution methods, fine details of the imaged structures become clearly visible even when they are not recognizable in the input images.
Original languageGerman
QualificationDoctorate / Phd
Awarding Institution
Supervisors/Advisors
  • Schweikard, Achim, Supervisor
  • Fischer, Bernd, Supervisor
  • Mertins, Alfred, Supervisor
Publication statusPublished - 2009

Cite this