Computer assisted systems for orthopaedic surgery use pre-operatively acquired 3D-data sets for the planning and assistance of surgery. The most frequently used image modality is the computed tomography (CT). Its advantages are geometrically correct images with high spatial resolution; although the patient is exposed to ionizing radiation during the acquisition. For this reason, the harmless magnetic resonance (MR) imaging is a better modality. A further benefit is the improved image contrast. The objective of this thesis is a completely MR-based orthopaedic surgery without using any CT-data. The biggest problem is the geometrical distortion of the MR-data sets. The distortions are originated in the different magnetic properties of the human tissue and the surrounding air, and the actual size of the distortions finally depends on the patient. Therewith, the application of MR-data sets in surgical navigation requires at least a geometrical correction of the data. The segmentation of the MR-data and the intra-operative registration on the patient are further problems. The concern of this thesis is to provide solutions for these three problems. Firstly, the MR-data is segmented for mainly two reasons. A determination between air and tissue is needed for the distortion correction. Moreover, a reference structure, the bone, is required for the pre-operative planning and intra-operative registration. If more than one data set is used, a prior registration of the data must be enforced. This is achieved by means of mutual information and Chi-square based registration algorithms with rigid transformation. The segmentation approaches use either separate 2D-slices of the data set or the complete 3D volume. The restriction of the user interaction to a minimum is one important feature. At the same time a simple possibility for the interaction with the algorithm is ensured. The main objective of this thesis is to analyse geometrical distortions and to find ways of their correction. A fast multigrid-based numerical approach is applied in order to compute the distribution of the magnetic field inside the MR-scanner. With this approach, the expected geometrical distortions are computed and analyzed. Further, approaches for the geometrical correction are developed and presented. The third main focus is the intra-operative registration of the pre-operative planning. Due to timing restrictions during surgery, the approach is divided into an initial pose estimation and a later optimization. During the initial estimation, the operator defines simple anatomical structures which are easy to find. A first registration can be performed with the help of these structures. For the improvement of the outcome, a further optimization is achieved by means of a fast octree-spline based approach.
«
Computer assisted systems for orthopaedic surgery use pre-operatively acquired 3D-data sets for the planning and assistance of surgery. The most frequently used image modality is the computed tomography (CT). Its advantages are geometrically correct images with high spatial resolution; although the patient is exposed to ionizing radiation during the acquisition. For this reason, the harmless magnetic resonance (MR) imaging is a better modality. A further benefit is the improved image contrast....
»