An accelerated framework for high-resolution X-ray holographic reconstruction

Area of Science:

• Physics
• Materials Science
• Biomedical Imaging

Background:

• X-ray propagation-based phase contrast imaging is crucial for high-resolution 3D structural reconstruction.
• Phase retrieval in this technique is computationally challenging, with existing software facing performance and hardware limitations.

Purpose of the Study:

• To develop a high-performance software solution, HiHolo, for holographic X-ray phase contrast imaging.
• To introduce and evaluate three improved iterative phase retrieval algorithms for enhanced reconstruction quality and efficiency.

Main Methods:

• Developed HiHolo software utilizing the CUDA-MPI architecture for parallel processing.
• Implemented three novel iterative phase retrieval algorithms: alternating projections with probe, extrapolation iteration, and parallel iterative reprojection.

Main Results:

• HiHolo demonstrated a 24%-37% performance improvement over existing software.
• Near-linear scalability was observed across multi-GPU systems.
• The parallel iterative reprojection method achieved a 6-14 times speedup for 3D reconstruction compared to serial processing.

Conclusions:

• HiHolo provides an efficient framework for high-quality holographic reconstruction in X-ray phase contrast imaging.
• The developed algorithms effectively reduce artifacts and enhance spatial resolution.
• The software offers significant performance gains and scalability for advanced imaging applications.