This is a proposal to install a state-of-the-art computer for the sole purpose of doing fast tomographic reconstructions of 3D micro-CT images. These images consist of 5123 to 10243 voxels and the tomographic reconstruction process currently requires from several days up to a week's computation using several powerful workstations in parallel. As our micro-CT scanner (supported through RR-11800) can scan specimens at a rate of one per day, this facility's throughput is severely limited by the 3D image reconstruction process. Moreover, currently the computers we use for reconstruction are needed for analysis and display of the reconstructed 3D image data so that the current reconstruction approach further reduces the data throughput. At present we have six NIH grant-supported intramural investigators who routinely use the facility as well as an additional three intramural investigators who are currently anxious to expand their use from a feasibility evaluation to routine use. Several extramural investigators have also expressed an interest in use of the facility. They have been excluded from routine access because of our throughput limitations. The requested computer is a Silicon Graphics Inc. "Origin 2000" computer which was selected after testing several other systems, putatively of similar performance specifications. The main factor taken into account was the duration of computing a 10243, 16 bit, voxel image from xray projection images (each 10242, 16 bit, pixels) obtained at 0.5 degree intervals around the specimen, the availability of 64 bit double precision floating point processing and the ease of programming the computer for optimal use. This approach of using a single, dedicated, computer is efficient in that the reconstruction process is essentially identical for all users, regardless of their image analysis interest. If each investigator were to do their own tomographic image reconstruction the total increase in compute capacity for the group would be more expensive than the proposed approach. Moreover, as new tomographic image reconstruction algorithms are implemented, this is greatly facilitated when only one computer need be reprogrammed.
|Effective start/end date||5/20/98 → 5/19/99|
- National Institutes of Health
Explore the research topics touched on by this project. These labels are generated based on the underlying awards/grants. Together they form a unique fingerprint.