Real-time markerless trocar pose acquisition with RGB-D sensor

There is a crescent demand of transparent training programs based on an objective metrics to measure surgical skill. A fundamental source of objective parameters is the real-time analysis of positions, paths and movements of laparoscopic tools. Unfortunately, the equipment required to acquire accurate spatial measurements is expensive and specialized. In this work we address the problem of laparoscopic tools tracking, focusing on trocar tool. Trocar is the essential tool to guarantee the access to the surgical area during laparoscopic surgery, which is the most common minimally invasive technique performed in clinically practice. Currently researchers have predominantly employed rather expensive magnetic [1] or video based [2,3] tracking. The main limitations of the first approach is that it requires to fix an external sensor to measure the position of the instruments, while the second type of methods suffer from limited precision. The availability of light structured depth sensor, such as Microsoft's Kinect, allows to address tracking problems with a very compact and cheap device. The main contribution of this work is a method for markerless trocar tracking based on mixed depth and color information.