Techniques in Iterative Proton CT Image Reconstruction

TL;DR

This review discusses physics models and iterative algorithms for proton CT image reconstruction, emphasizing methods to handle scattering effects and improve computational efficiency.

Contribution

It provides a comprehensive overview of modeling and iterative algorithms, including new flexible projection methods for proton CT reconstruction.

Findings

Projection algorithms efficiently solve large sparse systems

Block-iterative and string-averaging schemes enhance flexibility

Parallel implementations reduce computation time

Abstract

This is a review paper on some of the physics, modeling, and iterative algorithms in proton computed tomography (pCT) image reconstruction. The primary challenge in pCT image reconstruction lies in the degraded spatial resolution resulting from multiple Coulomb scattering within the imaged object. Analytical models such as the most likely path (MLP) have been proposed to predict the scattered trajectory from measurements of individual proton location and direction before and after the object. Iterative algorithms provide a flexible tool with which to incorporate these models into image reconstruction. The modeling leads to a large and sparse linear system of equations that can efficiently be solved by projection methods-based iterative algorithms. Such algorithms perform projections of the iterates onto the hyperlanes that are represented by the linear equations of the system. They perform these projections in possibly various algorithmic structures, such as block-iterative projections (BIP), string-averaging projections (SAP). These algorithmic schemes allow flexibility of choosing blocks, strings, and other parameters. They also cater for parallel implementations which are apt to further save clock time in computations. Experimental results are presented which compare some of those algorithmic options.