Robust Endoscope Motion Estimation Via an Animated Particle Filter for Electromagnetically Navigated Endoscopy

Electromagnetically navigated endoscopy, which is increasingly applied in endoscopic interventions, utilizes an electromagnetic sensor attached at the endoscope tip to measure the endoscope movements and to navigate the endoscope in the region of interest in the body. Due to patient motion and magnetic field distortion, sensor electromagnetic tracking (EMT) measurement inaccuracy and dynamic jitter errors remain challenging for electromagnetic navigation. This paper proposes a new tracking framework of an animated particle filter that integrates adaptive particle swarm optimization into a generic particle filter to significantly boost electromagnetic trackers. We validate our method on a dynamic phantom and compare it to state-of-the-art EMT methods. Our experimental results demonstrate the effectiveness and robustness of our method, which provides position and orientation accuracy of 2.48 mm, 4.69° that significantly outperformed previous methods at least with tracking error of 4.19 mm, 7.75°. The tracking smoothness was improved from 4.09 mm, 3.37° to 1.84 mm, 2.52°. Our method successfully tackled the particle impoverishment better than standard particle filters.