Beveled-tip needle-steering using 3D ultrasound, mechanical-based Kalman filter and curvilinear ROI prediction

Abstract

This paper introduces a new robust 3D ultrasound needle detection approach integrated in a 3D needle steering system associated to a real-time path planning. The robustness of an existing algorithm is improved by limiting the needle detection to a curvilinear region of interest (ROI) using a novel mechanical-based prediction model. This linear model is also used in a Kalman filter to reduce detection noise and reject false detections. These two improvements drastically increase quality of our feedback. Finally, the 3D needle steering system is able to reach a target in phantoms with a maximal error of 0.8 mm without obstacle and 1.6 mm with obstacle.