An Active Modeling Scheme for Deflection Prediction of Robotic-Assisted Flexible Needle

The needle puncture is a representative of minimally invasive surgery. However, the complex interactions between the needle and the tissue pose a challenge in capturing needle deflection during the puncture procedure. This paper develops an active modeling (AM) scheme to predict the needle tip position in real time. First, uncertainties due to needle-tissue interactions are lumped into a modeling error term and introduced as an extended state to reconstruct the state-space model. Subsequently, the unscented Kalman filter is selected as the estimator in AM to predict the needle tip position in real time. Finally, a series of comparative experiments on our self-built robotic-assisted flexible needle platform demonstrate the effectiveness and prediction accuracy of the proposed scheme.