Adaptive Mean-Shift Kalman Tracking of Laparoscopic Instruments

Abstract

Laparoscopic surgery becomes increasingly popular due to high benefits to both surgeon and patients. In this paper, we propose the adaptive mean-shift Kalman tracking algorithm based on the mean-shift algorithm and the Kalman filter for tracking a laparoscopic instrument in laparoscopic surgery. An iterative update of the target candidate in the mean-shift process can improve the tracking performance over a typical mean-shift algorithm. In addition, the Kalman filter is employed to enhance the chance of tracking accuracy, especially when the object disappears from the scene. In this study, we tested the tracking performance of our proposed algorithm by using the different situations from simulated videos. Our experimental results show that the proposed algorithm can locate the target object correctly even when the size and the shape of the target have been changed. In the difficult situation when the target is hiding behind an obstacle, this algorithm can still track the target object correctly after it becomes apparent. Therefore, this proposed algorithm can be used for locating the tip of the laparoscopic instrument in real laparoscopic surgery. the color tracking algorithm to control a robotic laparoscope instead of using human; however, this method cannot track many types of instruments. Casals et al. (3) introduced feature tracking algorithm based on shape information of a surgical instrument; however, it works only with a specific surgical instrument. Lee et al. (4) proposed a color and shape tracking algorithm by using the contour of the surgical instrument. This algorithm worked well in the normal situation, but not when the instrument is blocked by some obstacles. Wei et al. (5) presented a simple algorithm for tracking target features. However, this algorithm is based on the artificial color marks attached to a surgical instrument, but there are many disadvantages, such as sterilization of the mark on the surgical instrument and its convenience in the real practice. Because of some limitations in previous methods, in this paper, we propose a new object tracking algorithm to track the laparoscopic instrument called the adaptive mean-shift Kalman algorithm (6), which is based on the mean-shift algorithm (7) and the Kalman filter (8).In this algorithm, the size of the target candidate can be adjusted during tracking processes to increase the chance of tracking. We simulated videos with different scenarios to test the performance of our proposed algorithm. Furthermore, the proposed algorithm is intended to use for controlling our new laparoscopic-holder assistant robot (9-10) and tracking the tip of its instruments in laparoscopic surgery.