Remote manipulation during robot-assisted surgery requires proficiency in perception, cognition, and motor skills. We aim to understand human motor control in remote manipulation of robotic surgical instrument and attempt to measure motor skills. Three features, smoothness, normalized jerk score, and two-thirds power law coefficient, estimating the motor skills of surgeons were analyzed. These features were calculated through segments, extracted from continuous end-effector trajectories during suturing, knot-tying, and needle-passing surgical tasks, performed by 8 right-handed subjects on bench-top models using da vinci surgical kit control system. Each subject repeated each task five times. Totally 1567 segments were extracted, 413, 437, and 717 segments performed by experts, intermediates, and novice subjects, respectively. Dynamic change of kinematic properties was analyzed to evaluate the relationship between considered features and motor skill level. Results show smoothness is significantly correlated with normalized jerk score and both features are significant measures of expertise levels. Also, results show the power law is violated by many end-effector trajectories and there is no relationship between obeying two-thirds power law, smoothness and jerk. We conclude trajectory is improved from non-smooth and jerky in novices to smooth in expert surgeons. This property may be used for motor skill evaluation. It is unlikely that obeying two-thirds power law be a valid property of all end-effector trajectories. However, power law coefficient may be a discriminant feature for levels of expertise. The results are also applicable in a home-based monitoring platform, to monitor motor functioning improvement of stroke patients during rehabilitation process.
Automated robot-assisted surgical skill evaluation: Predictive analytics approach
