This Small Business Innovation Research Phase I project will demonstrate the potential for a one to two order of magnitude improvement in three-dimensional (3-D) image reconstructions by applying an innovative parallel-processing architecture to the NSF- developed VOIR? and other 3-D image reconstruction algorithms. By drastically reducing reconstruction time, the Image Matrix Processor (IMP) will remove a major obstacle in the development of high-resolution medical and industrial 3-D CT. The IMP operates on matrices the way an array processor operates on vectors. Designed to be installed in an expansion slot of a high-end workstation, the IMP makes it possible for the first time to perform rapid volumetric reconstructions and 3-D segmentations of large data sets, such as those generated by computed tomography (CT) or magnetic resonance imaging (MRI) systems. The IMP's architecture will also support the volumetric visualization and real-time manipulation of large 3-D images, providing an ideal desktop environment for scientific and medical applications of the future. The Phase I Program will do a step-by-step microcode simulation of an IMP processor running 3-D CT reconstruction algorithms to demonstrate that, using commercially available technology, an affordable IMP/workstation could be capable of performing 5123 reconstructions in less than 2-hours. A conceptual design of an IMP/workstation architecture capable of meeting this goal will be developed. The Phase III commercial goal is to offer an affordable, workstation-based capability (IMP/workstation) with the power to reconstruct 10243 images in less than 2 hours.
