Open Collections

Towards real-time tissue surface tracking with a surface-based extended kalman filter for robotic-assisted minimally invasive surgery

University of British Columbia

The use of registered intra-operative to pre-operative imaging has been proposed for many medical interventions, with the goal of providing more informed guidance to the physician. The registration may be difficult to carry out in real-time.Therefore, it is often necessary to track the motion of the anatomy of interest in order to maintain a registration. In this work, a surface based Extended Kalman Filter (EKF) framework is proposed to track a tissue surface based on temporal correspondences of 3D features extracted from the tissue surface. Specifically, an initial 3D surface feature map is generated based on stereo matched Scale Invariant Feature Transform (SIFT) feature pairs extracted from the targeted surface. For each consecutive frame, the proposed EKF framework is used to provide 2D temporal matching guidance in both stereo channels for each feature in the surface map. The 2D feature matching is carried out based on the Binary Robust Independent Elementary Feature (BRIEF) descriptor. If the temporal match is successful in both stereo channels, the stereo feature pair can be used to reconstruct the feature location in 3D. The newly measured 3D locations drive the EKF update to simultaneously estimate the current camera motion states and the feature locations of the 3D surface map. The framework is validated on ex vivo porcine tissue surface and in vivo prostate surface during a da Vinci radical prostatectomy. The peak and mean fiducial errors are 2.5 mm and 1.6 mm respectively. Compared to other methods, the surface based EKF framework can provide a reliable 2D feature matching guidance for each feature in the 3D surface map. This maintains a chance to relocate a feature that was lost for a significant period of time. Such a surface based framework provides persistent feature tracking over time, which is crucial to drift free surface tracking. With implementation on a Graphic Unit Processor (GPU), real time performance is achieved.

Text

2014-06-12

Attribution-NonCommercial-NoDerivs 2.5 Canada

http://hdl.handle.net/2429/46992

Electrical and Computer Engineering

University of British Columbia

UBCV

http://creativecommons.org/licenses/by-nc-nd/2.5/ca/