A reliable surface reconstruction system in biomedicine

For common biomedical imaging facilities, such as CT, MRI, and confocal microscopy, the acquired scans are sequential parallel sections. The object of interest in each section image can be extracted by segmentation procedure to form serial parallel planar contours. How to reconstruct a trustworthy surface from these contours is a crucial issue in biomedical 3D visualization. In this paper, we propose an automatic, fast, and reliable surface reconstruction system. An improved correspondence-determining algorithm is proposed in the system to provide more reasonable contour-correspondences than the existing algorithms. It can handle more general input data, and does not produce wrong reconstruction results. A hybrid tiling algorithm is presented to tile the corresponding contours without the requirement of a contour-matching procedure. It can also handle the branching problem without any modification. For degenerate cases and branches, intermediate contours are introduced by means of contour interpolation to enhance the reconstruction results. The surface area and volume are also calculated to facilitate the practical applications.