Publication Details
Jing Xu and Andrey Chernikov.
Published in VMASC 2013 Capstone Conference, pages 88 -- 93, Suffolk, VA, April, 2013
Abstract
The processing and understanding of medical images rely heavily on the ability to visualize the surfaces of the organs or the tissues represented in the images. This visualization capability is especially critical for three-dimensional images which are composed of a number of two-dimensional slices aligned and stacked on top of each other, such as the ones obtained with computed tomography, magnetic resonance, ultrasound, and other technologies. One of the most frequently used techniques for extracting the surfaces from medical images is the Marching Cubes algorithm. Given a required intensity threshold, this algorithm processes the volumetric pixels (voxels) in the image one at a time and constructs a number of triangles in the three dimensional space that approximate the isosurface corresponding to this threshold. The union of all triangles from all voxels yields a high accuracy polyhedral representation of the isosurface which is suitable for rendering on graphics hardware or for further processing. This extraction of isosurfaces is frequently used in interactive and/or iterative applications, and therefore the speed of extraction needs to be minimized. In this paper we present our parallelization of the Marching Cubes algorithm which achieves a nearly linear decrease in the processing time with respect to the number of used hardware cores..
