This Shared Instrumentation Grant is aimed at providing fast interactive graphics, computational power and memory resources, via a Silicon Graphics Onyx2 Infinite Reality visualization supercomputer to support the research of a group of investigators working on the processing, analysis and visualization of large, multi-dimensional medical images. All of the user projects are involved with the development and validation of computationally intensive, advanced computer vision, image analysis and image processing algorithms. The significant increases in processing, graphics and memory over the hardware currently available to this user group provided by the new system will speed algorithm execution time, visualization rendering time and avoid the use of sub- images and image sub-sampling currently necessary due to memory constraints. Such processing power will speed and enhance algorithm development and allow for interactivity while performing quantitative image analysis tasks for validation and clinical evaluation. The vastly enhanced graphics and computational power will allow dynamic 3D rendering and computational tasks previously infeasible. This system provides a single platform that can serve to graphically initialize and execute algorithms, visualize results, allow for correction of re- initialization, and finally, clinical interpretation and measurement. These projects involve computationally complex tasks such as optimizing multi-dimensional objectives and solving differential equations using finite element methods. The interactivity and graphics capabilities of the Onyx2, particularly in 3D surface and volume rendering, will greatly enhance the algorithm development and validation process. Two projects deal with cardiac image sequences and thus have need for the demanding graphics capabilities necessary for 2D and 3D rendering of temporal sequences with overlaid functional maps. Three other projects require 3D rendering of large brain volumes, with overlaid function or structural or structural surfaces. The Onyx2 would allow such rendering in an interactive fashion for algorithm evaluation and clinical interpretation, diagnosis and treatment planning.
