Computational x-ray imaging algorithms: enhancing current schemes and developing models for future schemes

Syed, Sajid Ali

doi:https://doi.org/10.21985/n2-fqy3-3426

Work

Computational x-ray imaging algorithms: enhancing current schemes and developing models for future schemes

Public

Download PDF

Download All Files (.zip)

As fourth-generation synchrotron x-ray sources start to come online, they bring with them the promise of increased imaging speed and resolution for current imaging modalities, while enabling entirely new ones. Employing the increased brightness at hard x-ray energies begins with having efficient focusing optics in place. Zone plates are popular choices for the task of hard x-ray focusing, particularly as developments in nanofabrication have led to the fabrication of high-aspect-ratio zone plates that deliver high efficiency and high resolution. However, due to the presence of waveguide effects, previous estimates for tilt misalignment sensitivity do not apply. In this thesis, we present a novel exploration of these limits based on the multislice model of x-ray wave propagation in inhomogeneous media, followed by experimental validation. Tomography is a widely used imaging scheme that combines multiple projections through a sample to generate a three-dimensional image. However, errors in experimental apparati generate artifacts in the final reconstruction, thereby presenting themselves as a serious obstacle to obtaining high resolution images. Combined with the higher data rates expected from increased brightness, this necessitates the development of robust and reliable high-throughput reconstruction tools that account for center-of-rotation errors. In this thesis we detail the development and performance of one such tool (PIRT) and demonstrate its capabilities. Finally, we explore the possibility of beyond-depth-of-focus imaging as an imaging scheme that increased brightness would enable. This scheme requires us to pay attention to within-sample diffraction effects. Within the current landscape of optimization-based reconstruction tools being developed for this, there is an urgent need to evaluate the scalability of x-ray wave propagation algorithms as teravoxel-sized images begin to be contemplated. In this thesis, we present novel implementations of distributed memory parallel x-ray wave propagation algorithms, and compare their scalability. The thesis concludes by offering suggestions on improvements to current computational image reconstruction algorithms, and a call to revive prior proposals for developing a community-based software toolkit for scattering-based imaging techniques.

Creator