Liver cancer is one of the leading causes of death worldwide. One of the reasons for that is the high tumor recurrence rate. The only way to reduce the recurrence rate is to ensure that all carcinogenic cells are destroyed after intervention. Unfortunately, the information available to assess the outcome of an intervention is limited. In the clinical routine, a pair of pre- and post-operatively gathered computed tomographies (CT) of the abdomen are typically compared to decide whether the patient needs further treatment. However, the post-operative liver will be deformed due to breathing and intervention which will complicate the comparison task by simple inspection of both images.

The results presented in this thesis will support the physician during the outcome validation process after minimally invasive interventions and open liver surgeries. Therefore, the physician is provided with qualitative measures and visualizations that support him in the decision making task. The basis of a reliable outcome validation is an accurate non-rigid registration method.

This thesis proposes to combine internal correspondences at vessel ramifications and landmarks at the surface of the organ to increase the accuracy of the registration results. The internal correspondences are the result of a novel efficient and fully automatic graph matching method. Landmarks at the surface of the liver are given by a method that detects the organs that are adjacent to it at each surface point. Both types of landmarks are incorporated in a FEM-based registration. The registration method has been tested in 25 pairs of pre- and post-operative clinical CT images achieving an average accuracy of 1.22 mm and a positive predictive value of 0.95.

In consequence of the accuracy obtained with the proposed methods the physician is able to determine with certainty if the outcome of the intervention was satisfactory. Hence, he can without delay decide to re-treat the patient if needed to remove the remnant tumor. This fast response could at the end reduce the tumor recurrence rate.