Automatic Centerline Extraction for 3D Virtual Bronchoscopy

Tsui Ying Law, Pheng-Ann Heng

Abstract: Centerline extraction is the basis to understand three dimensional structure of the lung. In this paper, an automatic centerline determination algorithm for three dimensional virtual bronchoscopy CT image is presented. This algorithm has two main components. They are end points retrieval algorithm and graph based centerline algorithm. The end points retrieval algorithm first constructs a binary tree which links up all necessary center points of each region in each slice from segmented lung airway tree volume data. Next, it extracts the end points of the lung airway tips from the binary tree. The graph based centerline algorithm reads the end points and applies distance transform to yield a distance map which shows all shortest paths from the start point to those end points. Then, modified Dijkstra shortest path algorithm is applied in the centerline algorithm to yield the centerline of the bronchus. Our algorithm is tested with various CT image data and its performance is encouraging.