Reconstruction for time-domain in vivo EPR 3D multigradient oximetric imaging--a parallel processing perspective

Area of Science:

• Biophysics
• Medical Imaging
• Computational Science

Background:

• Three-dimensional (3D) oximetric Electron Paramagnetic Resonance (EPR) imaging provides vital information on tissue oxygenation and can differentiate normal from tumor tissues.
• Acquiring and processing large datasets (gigabytes) from 3D EPR oximetry is computationally intensive, involving filtering and 3D Fourier reconstruction, which is slow on uniprocessor systems.

Purpose of the Study:

• To develop and evaluate a parallel computing framework to accelerate computationally demanding 3D EPR oximetry imaging processes.
• To significantly reduce the processing time for filtration and 3D Fourier reconstruction tasks in 3D oximetric EPR imaging.

Main Methods:

• A parallelization framework was implemented using OpenMP runtime support and parallel MATLAB.
• A parallel C++ code was developed using OpenMP and executed on multi-core AMD Opteron processors for filtration.
• A parallel MATLAB code was created for 3D Fourier reconstruction and oximetry computation.

Main Results:

• The parallel filtration code achieved a speedup factor of 46.66 compared to serial MATLAB code.
• Parallel 3D Fourier reconstruction and oximetry computation yielded speedup factors of 4.57 and 4.25, respectively.
• The parallel implementation significantly reduces computational burden, enabling near real-time acquisition of biophysical parameters.

Conclusions:

• Parallel computing, utilizing OpenMP and parallel MATLAB, effectively accelerates 3D EPR oximetric imaging.
• This approach provides high computational power for rapid analysis of biophysical parameters from 3D EPR oximetry.
• The developed system is accessible and offers near real-time capabilities for advanced biomedical research.