Centerline Extraction Based on Hessian Matrix and Scale Space Analysis

In order to extract the medial axes of vascular objects from 3D medical volume data, a precise centerline extraction algorithm is proposed. A binary image sequence which has been segmented is taken as the input data. Distance transform is performed for the data to generate a distance map. Based on the distance map and for each target-voxel which belongs to the cavity of vascular structure, a Hessian matrix is computed. By judging the eigenvalues and eigenvectors of Hessian matrix for each target-voxel, a coarse centerline is extracted according to some rules. After modifying all points in the coarse centerline by using scale space analysis, a precise centerline will be obtained. The experimental results illustrate the simplicity of Hessian matrix extraction and the precision of scale space analysis. And the whole centerline extraction algorithm has been applied to 3D medical volume data visualization-virtual endoscopy.