Landmark-based surgical navigation

Navigational support is a widely adopted technique in surgery that has become a part of the clinical routine. This navigation support either comes in the form of an abstract display that shows for example the distance and direction to the target position or in the form of augmented reality where segmented anatomical structures of interest are overlaid onto a visual image sequence in real-time. In this paper we propose a cost-effective real-time augmented reality approach using an off-the-shelf tablet PC in combination with a novel 2D/3D point correspondence mapping technique. The proposed point pattern matching algorithm is tailored towards moderate projective distortions and suitable for computational low-power devices. Experiments and comparisons were done on synthetic images and accuracy was measured on real scenes. The excellent performance is demonstrated by an Android 3D guidance application for a relevant medical intervention.