Visual Tracking of Laparoscopic Instruments

In order to enable the standard set-up used in the training and the practice of minimally invasive surgery procedures to be more immersive, it is necessary to design and develop an interactive surgeon computer interface. In this paradigm, visual tracking of the surgical instrument takes the center stage in design of such an interface. In this paper, we present a novel 3D visual tracking system using a single endoscopic camera. 2D feature recognition and stochastic approach are applied to assist the tracking system and the 3D reconstructed model of instrument is also provided. Two tracking algorithms based on standard Kalman filter and extended Kalman filter are proposed and compared in the article. The performance of the proposed tool tracking algorithms are evaluated and discussed using both emulated and video footage of the actual surgical environment. The results show our methods can track the objective well in “Good” quality scenes and those scenes with strong contrast between tool and background. 