Electron Paramagnetic Resonance (EPR) is a spectroscopic technique that detects and characterizes molecules with unpaired electrons (i.e., free radicals). Although it is closely related to nuclear magnetic resonance (NMR) spectroscopy, EPR is still under development as an imaging modality. Unlike other imaging modalities, EPR is able to take direct measurements of tissue oxygen concentration in a manner that is not dependent on complex biological processes such as ligand binding specificity or tracer metabolism. The single-point imaging (SPI) scheme is essentially a phase-encoding technique that operates by acquiring a single data point in the free induction decay (FID) after a fixed delay (the phase encoding time), in the presence of static magnetic field gradients. SPI produces artifact-free images because it does not measure the time evolution of the magnetization. The goal of this project is to provide computational methodology and resources that will advance the state of the science in EPR. A particular focus of this collaboration is the development of reconstruction methodology that will improve the quality of oximetric images obtained using the SPI technique.
Advances to the time-domain EPR instrument which were made in a previous year by Dr. Sankaran Subramanian and Mr. Nallathamby Devashayam of RBB/NCI have improved the feasibility of in vivo imaging using the time-domain, radiofrequency, single-point technique. These advancements have generated interest in the prospect of deploying a dedicated computational cluster. HPCIO made considerable progress in FY 2006 in re-engineering a piece of software that filters and removes noise and background from the free-induction decay (FID) signals. The time required to process the FIDs has been reduced from over 2 hours in the original Matlab code to less than 30 seconds on 4 AMD Opteron processors. Progress has also been made on re-engineering the software that computes and displays the oxygen maps. In FY 2006, a shared-memory, high-performance computing system was specified ordered. The goal is to deploy the re-engineered software onto this dedicated parallel computing system.
Current and Future Work
In FY 2007, we can anticipate the arrival of the 4-processor, dual-core dedicated parallel computing system that was ordered in FY 2006. Since this system will be located on floor B2.5 of building 10, space has already been identified for the system. Therefore, there will be no delay in deploying the system or the software. Some modification to the FID-processing software may be necessary to accommodate the dual-core architecture of the new parallel computing system. Extensive modifications and explicit parallelization are likely required for the oxygen-mapping software. That software will likely be parallelized using the Matlab Distributed Computing Engine.
- Murali Krishna Cherukuri, Ph.D., Radiation Biology Branch, NCI
- James Mitchell, M.D., Ph.D., Radiation Biology Branch, NCI
- Sankaran Subramanian, Ph.D., Radiation Biology Branch, NCI
- John Cook, Ph.D., Radiation Biology Branch, NCI
- Nallathamby Devashayam, Radiation Biology Branch, NCI
S. Subramanian, C.A. Johnson, N. Devasahayam, K. Matsumoto, F. Hyodo, J. Cook, and M.C. Krishna, “In vivo Spectral-Spatial Imaging for Oxygen Mapping Using Single-Point, Time-Domain Electron Paramagnetic Resonance,” Proceedings of Cambridge Healthtech Institute’s Second Annual Conference on In Vivo Molecular Imaging, San Diego, CA, IMG536, 2005.
S. Subramanian, C.A. Johnson, N. Devasahayam, K. Matsumoto, F. Hyodo, J. Cook, and M.C. Krishna, “In vivo Spectral-Spatial Imaging for Oxygen Mapping Using Single-Point, Time-Domain Electron Paramagnetic Resonance,” Proceedings of the 2006 IEEE International Symposium On Biomedical Imaging, Crystal City, VA (2006), SA-PM-OS4.1.
- Size of job (number of gradient measurements): 25×25×25x3
- Original cost of processing the FIDs: 40 minutes
- Cost of processing the FIDs on 4 processors (re-engineered): 30 seconds
- Speedup factor: 80