Kalman filter based vision-servoing alignment system for human ear surgery

Abstract

To treat a worldwide common ear disease called Otitis Media with Effusion (OME), occurring in adults and children, a robotic device allowing fast and automatic grommet tube insertion has been designed in our earlier works. However, the previous instrument has to be manually placed close to the tympanic membrane before the insertion procedures. For the purpose of realizing a fully automated surgical process, locating a suitable insertion path from outside the ear and controlling the working channel to be aligned to the axial direction of auditory canal from the fiberscopic vision forms a crucial step. In this paper, a fast and robust vision based servomechanism is proposed to achieve the planar alignment. An image segmentation algorithm based on the phase information and combining with the Hough Transformation technique is proposed to detect and track the region of auditory canal through the image sequences. With the help of the linear encoder, a Kalman filter is introduced to estimate the motion states employing the visual and motion measurements simultaneously to construct a new look-and-move servomechanism. Different from the traditional look-then-move approach, in which the low-speed image processing rate may dramatically affect the control performance, the proposed framework can predict the unmeasured motion states through the recursion of Kalman filter, resulting a more effective and robust servo control performance with smooth and well-aligned trajectories.