An Adaptive Unscented Kalman Filter for Needle Steering with Missing Measurements

Abstract

In the robot-assisted puncture surgery, the measurements from the ultrasound image may be lost due to the uneven distribution of image grayscale and blurred image which could affect the estimation accuracy of the filter. Furthermore, the probability of the missing measurement cannot be precisely known due to the heterogeneity of biological tissue and the complexity of the surgery environment. Aiming at these problems, an adaptive unscented Kalman filter algorithm based on virtual measurement noise is proposed to estimate the pose of the needle tip in this paper. The missing measurement is converted into a virtual measurement noise which has an indefinite variance. Then the variance of the virtual noise is estimated in real-time to suppress the influence of missing measurements during the puncture process. Furthermore, according to the strong tracking filtering algorithm, an adaptive fading factor is constructed to reduce the sensitivity of the filter to the statistical characteristics of the noise. Finally, the proposed filter is applied to estimate the pose of the puncture needle and the effectiveness of the proposed method is verified via simulation experiments.