University of Utah Interactive, GPU-Based Level Sets for 3D Segmentation 
Aaron E. Lefohn, Joshua E. Cates and Ross T. Whitaker 
Scientific Computing and   Imaging Institute

Tumor Segmentation using our system Expert hand segmentation


    While level sets have demonstrated a great potential for 3D medical image segmentation, their usefulness has been limited by two problems. First, 3D level sets are relatively slow to compute.  Second, their formulation usually entails several free parameters which can be very difficult to correctly tune for specific applications.  The second problem is compounded by the first.  This paper presents a tool for 3D segmentation that relies on level-set surface models computed at interactive rates on commodity graphics cards (GPUs).  The mapping of a level-set solver to a GPU relies on a novel mechanism for GPU memory management.  The interactive rates for solving the level-set PDE give the user immediate feedback on the parameter settings, and thus users can tune three separate parameters and control the shape of the model in real time.  We have found that this interactivity enables users to produce good, reliable segmentations.  This paper presents qualitative and quantitative results from this tool on brain tumor segmentation to support this observation.
Aaron Lefohn University of Utah
Joshua Cates  University of Utah
Ross Whitaker University of Utah
"Interactive, GPU-Based Level Sets for 3D Segmentation" in Medical Image Computing and Computer Assisted Intervention" (MICCAI) 2003.   pdf  
Detailed description of the GPU-based sparse-grid level-set solver.  pdf  
University of Utah School of Computing Technical Report  pdf ps
MICCAI 2003 Presentation  ppt
A user segmenting a brain tumor from a 256 x 256 x 124 MRI   mpeg  quicktime
A user segmenting the cerebral cortex from a 128 x 128 x 88 MRI   quicktime