Jan Strehlow

• Minimal invasive therapies for treating liver tumors heavily depend on imaging. Motion in between the acquisition of different images and in between image acquisition and therapy action needs to be compensated to reach the therapy goal. This thesis presents motion compensation techniques for liver tumor therapies and addresses unsolved problems in their transfer to clinical practice. A first problem is motion in between the time-points of dynamic contrast-enhanced (DCE) magnetic resonance imaging time series, which is a prominent tool for imaging liver anatomies in therapy planning and assessment. We present a motion compensation employing a pipeline of liver segmentation, rigid preregistration and deformable registration to robustly compensate large differences in the respiratory state. The transfer of DCE motion compensations to clinical practice is hampered by the lack of homogeneous, direct, quantitative measures for motion compensation quality. We address this problem with a novel landmark annotation scheme that allows for direct sampling of ground truth on numerous cases with comparably low effort. In a second part, we address motion compensation problems during the intervention at the example of a novel, motion-compensated focused ultrasound (FUS) treatment system. We present a ready-to-use treatment system employing clinically approved hardware, in which motion-compensation is realized by observing motion several times per second, predicting the target motion, and finally adopting the treatment plan accordingly. Monitoring of treatment effect in the moving liver and in real-time is enabled by a multi-baseline thermometry approach employing a novel fast lookup. To prepare for regulatory approval of the use in a clinical study, clinical requirements for safety and efficacy are defined and their fulfillment is tested via a dedicated suite of experiments. The results suggest that safe and effective motion-compensated FUS is possible with the proposed system.