A Hybrid Method for Real-Time Bronchoscope Tracking Using Contour Registration and Synchronous EMT Data

Abstract

Background: Bronchoscopy is a difficult procedure for physicians to relate CT slices to bronchoscopic video images and maneuver the bronchoscope inside the airway tree. CT-guided bronchoscopy systems have been developed in the last decades to help physicians maneuver the bronchoscope inside the airway tree in a fast and precise way. Objectives: We aimed to develop a continuous guiding method for bronchoscopy with high tracking accuracy by matching bronchoscopy image contours with CT contours, and speed it up by using synchronous electromagnetic tracker (EMT) data, and to evaluate it on airway phantom with simulated respiratory motion. Materials and Methods: This method works based on two approaches combined together: 1) Contours detected in real bronchoscopy images and finding their equivalents in CT space. Contours are detected by a fast algorithm and CT contours are mapped by them in a perspective scheme. 2) EMT data which is used in a frame by frame approach to approximate the position of bronchoscope compared to its previous position. This differential approach causes a small search space and as a result higher tracking speed. The novelty of this work is using bronchoscopy image contours instead of the whole image combined with synchronous EMT data. This approach causes faster tracking and there is no need for landmark selection or centerline consideration before performing the main bronchoscopy. Results: The experimental results of implementing the proposed method show that this method can track the bronchoscope continuously. For evaluating the accuracy and robustness of tracking, virtual bronchoscopy images were generated at each frame position reported by the method and compared to corresponding real bronchoscopy image using mutual information. The experimental results present that this method can track a bronchoscope accurately and robustly in 96.3% of frames. Conclusion: Using contours instead of the whole image for registration can provide a continuous real-time bronchoscopy tracking procedure. Using EMT data in differential mode makes the proposed method robust to simulated respiratory motion.