In this paper, we describe an algorithm for recognizing the surgical tasks conducted during robotic microsurgery, especially for suturing a micro-blood vessel. In the proposed method, the surgical tasks are recognized using a hybrid HMM/RF model. The motion of the robotic system during each surgical task is modeled using a random forest, and the transition between the surgical tasks is modeled using a hidden Markov model, which considers the minimum duration of each surgical task in the hybrid HMM/RF model. The proposed method was evaluated by recognizing 12 tasks during the suturing of an artificial blood vessel using the microsurgical robotic system, and the tasks were successfully recognized to have a 78.5% accuracy.
