Proton stopping power images from Monte Carlo simulated dual-energy CT scans

TL;DR

This study demonstrates the feasibility of deriving proton stopping power images from Monte Carlo simulated dual-energy CT scans, showing potential for improved accuracy and reduced artifacts in tissue characterization.

Contribution

It introduces a Monte Carlo simulation-based dual-energy CT method for calculating proton stopping power, optimizing monochromatic energies to minimize residual errors.

Findings

Modest error and noise reductions in phantom inserts.

Visible reduction in beam-hardening artifacts.

Feasibility confirmed for DECT-based proton stopping power determination.

Abstract

We test the feasibility of calculating proton stopping power (SP) from a set of virtual monochromatic (VM) CT images, created with Monte Carlo-generated CT scans, using a dual-energy CT-to-SP procedure. The Monte Carlo CT simulations, with 70 and 150 kVp spectra, were modeled on the x-ray tube parameters of the SOMATOM Force dual-source DECT scanner and the Gammex RMI 467 tissue calibration phantom. Reconstructed SP images were created with two VM x-ray images, and the monochromatic energy pairs were chosen to minimize the SP residual errors using the root-mean-squared-error (RMSE). The contrast of phantom inserts has also been examined, in addition to relative errors of the average reconstructed insert values (which are calculated using the known electron densities and chemical compositions of each material, provided by the manufacturer). Results were also compared to a similar dual-energy-based SP conversion procedure for comparison. Modest reductions in errors and noise were seen in almost all inserts in both low- and high-density configurations of the phantom, with a visible reduction in beam-hardening streaking artifacts. Our Monte Carlo simulated DECT scans confirm the feasibility of previously proposed methods of DECT-based $ S(\rho_e, Z) $-determination. By utilizing VM images and only optimizing the RMSE over specific inserts, the residual error across the most vital bodily tissues may be reduced when extremely low- or high-density tissues are known to be present.