A.F.D. Ornelas, J. Lipp, C. Bouman, G. Buzzard, M. Ostendorf, V. Roman, J. Roth, D. Kisailus, S. Farajollahi
AFRL,
United States
Keywords: Image reconstruction, Noise reduction, Optimization, Inverse problems, Tomography, Phantoms, Computational modeling, Filtered Back Projection, Keratin, Hair
Summary:
Computed Tomography (CT) has revolutionized imaging by providing detailed insights inside materials nondestructively. The process of generating these images from raw X-ray data is known as CT reconstruction, with methods such as Filtered Back Projection (FBP), Iterative Reconstruction (IR) offering unique advantages. MBIR provides superior image quality by iteratively modeling the reconstruction until convergence. This study explores the application of Model-Based Iterative Reconstruction using Purdue’s MBIRJAX in the detailed analysis of wool fibers at the nanoscale. MBIRJAX is a Python package written by Charlie Bouman and Gregery Buzzard which implements Vectorized Coordinate Descent to update pixels in parallel to leverage GPU acceleration. This enables MBIRJAX to offer high-performance, rapid convergence, and robust reconstruction capabilities for tomographic data. Biologically self-assembled structures contain nano features that defines the functional properties of biomaterials. Here, by applying MBIRJAX to the analysis of wool fibers, we demonstrate its efficacy in resolving fine structural details that are otherwise challenging to capture with conventional imaging techniques. The results highlight the potential of MBIRJAX to enhance our understanding of complex architected material systems, paving the way for new innovations.