Optimal particle type and projection number in multi-modality relative stopping power acquisition

TL;DR

This study identifies helium ions as the optimal particle type and three projections as the ideal number for accurate, low-dose proton radiography combined with CT to improve cancer treatment precision.

Contribution

The paper introduces a novel optimization framework for selecting particle type and projection number in multi-modality radiography, demonstrating improved accuracy and reduced dose.

Findings

Helium ions with three projections yield the lowest RSP and range errors.

Using three projections achieves mean RSP error below 0.7% and range errors below 1 mm.

Alpha particle radiography delivers lower dose than X-ray radiography.

Abstract

Proton radiography combined with X-ray computed tomography (CT) has been proposed to obtain a patient-specific calibration curve and reduce range uncertainties in cancer treatment with charged particles. The main aim of this study was to identify the optimal charged particle for the generation of the radiographies and to determine the optimal number of projections necessary to obtain minimal range errors. Three charged particles were considered to generate the radiographies: proton (p-rad), helium ions ({\alpha}-rad) and carbon ions (c-rad). The problem formulation was viewed as a least square optimization, argmin_x (||Ax -b||_2^2 ), where the projection matrix A represents the cumulative path crossed by each particle in each material, b is the charged particle list-mode radiography expressed as the crossed water equivalent thickness (WET) and x is the relative stopping power (RSP) map to be optimized. The charged particle radiographies were simulated using the Geant4 Monte Carlo (MC) toolkit and an anthropomorphic adult head phantom. A was determined by calculating the particles trajectory through the X-ray CT employing a convex hull detection combined with a cubic spline path. For all particle types, the impact of using multiple projections was assessed (from 1 to 6 projections). In this study, tissue segmentation methods were also investigated in terms of achievable accuracy. The best results (mean RSP error below 0.7% and range errors below 1 mm) were obtained for a-rad when 3 projections were used. The dose delivered to the phantom by a single {\alpha}-rad was 8 \muGy, lower than the dose of X-ray radiography.