Tracking of Rigid Bodies in the Context of Minimally Invasive Surgical Training

Abstract

Minimally Invasive Surgery (MIS) poses additional challenges for the surgeon caused by the loss of direct contact with the surgical site, altered visual perception, motion inversion and scaling. Therefore, haptic guidance could be used to reduce these problems as it can prevent instrument intrusion into predefined areas, offers collision avoidance and eases the adaption to the system. An important prerequisite for this is the perception of relevant objects in the environment and the determination of their pose. In this work, a rigid Multi-Body Multi-Modality Multi-Camera 3D Tracker (M3T) is applied as a supportive tool for the training of surgeons in minimally invasive surgery. This includes the development of a manual detector for a more convenient initialization of the tracked objects and the introduction of a multi-region functionality for the M3T. Here, the object's model can be adapted to contain multiple regions. With this, occlusions, neighboring objects or sub-parts of the tracked object are considered throughout the tracking process. The effectiveness of this extension is then shown by the evaluation of the approach's accuracy, robustness and computational cost.