Muscle Fatigue Induced Hand Tremor Clustering in Dynamic Laparoscopic Manipulation

Differentiating muscle fatigue induced hand tremor of surgeons into different discernible levels is important in laparoscopic surgery. Systematic clustering can be used as a method to assess the risk of hand tremor which can largely affect the surgical performance. The prime challenges lying here are the detection of fatigue onset and classification of fatigue induced tremor level in dynamic laparoscopic tool manipulation. Conventionally, muscle fatigue is assessed with frequency domain analysis of the surface electromyography (sEMG) signal, where the detection process is predominantly valid only for isometric contraction of muscles. Conventional methods cannot be used for assessment of fatigue level in case of dynamic activities as the task itself modulates the frequency content of the myoelectric response. In this paper, we have proposed a novel polynomial Hammerstein model-based clustering of fatigue induced tremor, employing sEMG, and joint torques. The sEMG signal, containing muscle fatigue information, gets fused in this model dynamically through a Kalman filter. Model parameter-based clustering of the fatigue induced tremor level was implemented on eight subjects. Optimal number of cluster centers were found to be appropriately coherent with the fatigue inducing task epochs of the experiment. In spite of the subjective variations, the model parameter-based clustering method was able to differentiate among the fatigue-inducing tasks for all the subjects. We have concluded that this model-based clustering can successfully differentiate between different levels of the fatigue induced tremor in dynamic activity of laparoscopic tool manipulation.