Proton Computed Tomography: A Case Study for Optimal Data Acquisition

This study enables a comparison between two proposed algorithms in the literature for proton computed tomography (pCT) reconstruction; these are the analytical filtered back-projection (FBP) and the iterative algorithm of diagonally relaxed orthogonal projections (DROP) with total variation superiorization (TVS). The analytical model of the cubic spline path (CSP) has been implemented into DROP-TVS algorithm to account for scattering, using measurements of individual proton positions/trajectories and energy before and after traversing a selected phantom. Spatial resolution and relative stopping power (RSP) accuracy of the reconstructed images were used as comparison criteria. Parametric changes of projection angular ranges from 0
⁰
to 180
⁰
and 0
⁰
to 360
⁰
, as well as angular steps of 0.5
⁰
and 1
⁰
have been investigated. Experimental data of the CT0404 "Sensitom" and the CTP528 "Line Pair" modules of the Catphan® 600 phantom were used. DROP-TVS appeared to provide better noise reduction compared to FBP while the resolution worsened with reduced projections. On the other hand, FBP resulted in degraded image quality with more noise and worse spatial resolution compared to that of the DROP-TVS scheme. Interplane artifacts were present with both algorithms which became more acute in FBP reconstructions and with a limited number of projections. Reduced number of projections had an adverse effect on both reconstruction methods. The study outcomes highlight the optimal acquisition for pCT reconstruction for superior image quality but there must always be a trade-off between noise and spatial resolution for optimal data acquisition.