Development of a Robust Registration Method for Computer Assisted Surgery

Current developments in surgery and computer science expose many new technologies and techniques. In the context of minimal invasive surgery, the surgeon is using very small instruments, which are inserted into patient's body through ports. The disadvantage of new surgical approaches like minimal invasive techniques is the more difficult intraoperative orientation for the surgeon, based on intraoperative sensors that deliver only a small part of the intraoperative situation. To assist the surgeon also preoperative data will be acquired. Those data can be used in combination with techniques of augmented reality to enhance the field of view of the surgeon. Registration aims to close the gap between the preoperative operating planning scenario and the intraoperative support by calculating the transformation that maps one data set onto the corresponding other. In the underlying medical scenario, the developed registration algorithm calculates the transformation that maps preoperative data onto intraoperative data. The main idea is not to calculate the transformation in a traditional way by building point correspondences on both data sets based on minimal distances, like the Iterative Closest Point-algorithm. The developed algorithm instead characterizes preoperative and intraoperative surface data to build the desired correspondences between both data sets. Those characteristics can be created for vertices for example and encodes the local neighborhood around such vertices. The integration of additional information like directed characteristics or histological information is elementary and was implemented to enhance the accuracy and performance of the algorithm