Christian Schumann

• This PhD thesis describes methods for the support of the planning process for minimally- and non-invasive tissue ablation procedures. It focuses on the utilization of visualization and heuristic modeling to solve complex problems such as access path determination for needle-based tumor ablation therapies or sonication planning for high-intensity focused ultrasound. For needle-based interventions, the manual selection of suited access paths is addressed, which typically requires the careful inspection of every candidate path as a whole to make sure that no structure at risk would be penetrated. Especially for angulated paths that are not included in one image slice, this results in repeated inspection of multiple slices. To improve this process, a visualization method for the highlighting of infeasible paths in common 2D viewers is proposed. As an alternative to manual planning of percutaneous procedures, heuristic approaches to therapy plan optimization are investigated. An algorithm based on projection utilizing the GPU and image processing is proposed. Under consideration of multiple clinically relevant criteria, the method generates a list of optimal paths within a few seconds. A second, semi-automatic method is proposed, which represents a compromise between fully automatic and manual planning. It combines the fast projection based method with a numerical approach for Pareto-front approximation and allows for interactive exploration of the solution space. To improve the manual planning of high-intensity focused ultrasound therapies, a real-time approximation of temperature and thermal dose is developed. It is based on numerical simulation for a range of exemplary configurations in a preprocessing step. During interactive planning, the temperature or thermal dose field for the situation at hand is interpolated and combined with an approximation of the heat sink effect resulting from vessels in proximity. The result is visualized in an interactive manner.