Focused Registration of Tracked 2D US to 3D CT Data of the Liver

The paper deals with the registration of pre-operative 3DCT- data to tracked intra-operative 2D-US-slices in the context of liver surgery. To bring such a method to clinical practice, it has to be fast and robust. In order to meet these demanding criteria, we propose two strategies. Instead of applying a time-consuming compounding process to obtain a 3D-US image, we use the 2D-slices directly and thereby drastically reduce the complexity and enhance the robustness of the scheme. Naturally, the surgeon does not need the same high resolution for the whole liver. We make use of this fact by applying a focusing technique to regions of special interest. With this, we reduce the overall amount of data to register significantly without sacrificing the accuracy in the ROIs. In contrast to other attempts, the high resolution result in the ROI is combined in a natural way with a global deformation field to obtain a smooth registration of the whole liver. Overall we arrive at a method with a favorable timing. The proposed algorithm was applied to four different patient data-sets and evaluated with respect to the reached vessel-overlap on validation slices. The obtained results are very convincing and will help to bring non-linear registration techniques to the operation theater.