Accelerating fDOT image reconstruction based on path-history fluorescence Monte Carlo model by using three-level...

Area of Science:

• Biomedical Optics
• Computational Imaging
• Medical Physics

Background:

• Fluorescence diffuse optical tomography (fDOT) is a powerful imaging technique.
• Image reconstruction in fDOT, particularly using the path-history fluorescence Monte Carlo model, is computationally intensive and time-consuming.
• This excessive time requirement limits the practical application of fDOT.

Purpose of the Study:

• To develop and present a method for accelerating fDOT image reconstruction.
• To overcome the time limitations associated with the path-history fluorescence Monte Carlo model in fDOT.
• To enable efficient and practical fDOT imaging.

Main Methods:

• Implementation of a three-level parallel architecture.
• Leveraging distributed computing (cluster nodes), multi-core CPUs, and multi-processor GPUs.
• Selective utilization of GPU memories to minimize data-writing time and inter-iteration data transport.

Main Results:

• Demonstrated significant acceleration of fDOT image reconstruction.
• Validated the method's efficiency on general-purpose computing platforms.
• Confirmed the practical feasibility of using the path-history fluorescence Monte Carlo model for accelerated fDOT.

Conclusions:

• The proposed parallel architecture effectively accelerates fDOT image reconstruction.
• This method provides a practical solution for reducing computational time in fDOT.
• It enhances the utility of the path-history fluorescence Monte Carlo model for real-time or near-real-time fDOT imaging.