Reliable Fly-Throughs of Vascular Trees

Centerline Extraction of curvilinear objects is a crucial component of virtual endoscopy (VE), because it provides a path planning for both automatic and interactive navigation. During navigation, centerlines must be connected, smooth, and centered. Preserving connectivity was always a problem to many skeletonization methods. One such recent tempting method, is the voxel coding algorithm that extracts centerlines as a set of connected voxels inside the object. However, centeredness, smoothness, and even connectedness can be still missing in the nearby of branching nodes, especially tree structures with variable cross sections such as blood vessels. In this paper, we present a guided voxel coding (GVC) algorithm that modifies and extends the original method to extract reliable centerlines, suitable for vascular fly-throughs applications. Our proposed method works as follows, potential centerline voxels with maximal balls inscribed in the object are extracted and stored in a priori guidance map. Centerlines are initially extracted as trajectories inside the object, and then centered using a new mechanism that exploits the valuable information contained in the guidance map. A new thinning method is introduced to guarantee one voxel wide centerlines with no 2D manifolds or 3D self intersect. Finally, we build a tree graph of centerlines that provides useful information during navigation, especially, when it is required to find the shortest centerline path between two user defined points inside the object. The proposed algorithm is completely automated, computationally feasible and generates centerlines that are centered, connected, one voxel wide, smooth, and less sensitive to noise. Experimental results on different clinical data sets show the efficiency and robustness of the proposed method. Aly A. Farag and M. Sabry Hassouna are with the Computer Vision and Image Processing Laboratory, Department of Electrical and Computer Engineering, Room 415, University of Louisville, Louisville, KY, 40292, USA. (e-mail:farag, [email protected]). Robert Falk is with the Department of Medical Imaging, Jewish Hospital, Louisville, KY, 40202, USA. (e-mail:[email protected]). Stephen Hushek is with the MRI Department, Norton Hospital, Louisville, KY, 40202, USA. (email:[email protected]).