Experimental comparison of clinically used ion beams for imaging applications using a range telescope.

In particle therapy, the x-ray based treatment planning converting photon attenuation values to relative stopping power ratio (RSP) introduces clinically relevant range uncertainties. Recently, novel imaging technologies using transmission ion beams have been investigated to directly assess the water equivalent thickness (WET) of tissue, showing improved accuracy in RSP reconstruction, while potentially reducing the imaging dose. Due to their greater availability, protons have been mostly used for ion imaging. To this end, in this work, the influence of three ion species (protons, helium and carbon ions) on the image quality of radiographic WET retrieval has been explored with a dedicated experimental setup and compared to Monte Carlo (MC) simulations. Three phantom setups with different tissue interfaces and features have been irradiated with clinically validated proton, helium and carbon ion pencil beams under comparable imaging dose and beam settings at the Heidelberg Ion-Beam Therapy Center. Ion radiographies (iRADs) were acquired with an integration mode detector, that functions as a range telescope with 61 parallel plate ionization chambers. For comparison, experiments were reproduced in-silico with FLUKA MC simulations. Carbon ions provide iRADs with highest image quality in terms of normalized root mean square error, followed by helium ions and protons. All ions show similar capabilities of resolving WET for the considered phantoms, as shown by the similar average relative error < 3%. Besides for the slab phantom, MC simulations yielded better results than the experiment, indicating potential improvement of the experimental setup. Our results showed that the ability to resolve the WET is similar for all particles, intrinsically limited by the granularity of the detector system. While carbon ions are best suited for acquiring iRADs with the investigated integration mode detector, helium ions are put forward as a less technical challenging alternative.

Analytical simulator of proton radiography and tomography for different detector configurations.

PURPOSE: An analytical simulator of ion Radiography (iRad) and Computed Tomography (iCT) for protons is proposed to serve as imaging benchmark for different detector configurations. METHODS: The analytical simulator is applied to an anthropomorphic phantom and provides iRad and iCT benchmarks. Proton trajectories are traced relying on the Most Likely Path (MLP) algorithm. To simulate the proton trajectories the Multiple Coulomb Scattering (MCS) model embedded in the MLP algorithm is extended to non-uniform water equivalent materials according to variable substitution in the well-known statistical description in uniform water. The proton trajectories are instead estimated relying on the typical assumption of uniform water, thus causing intrinsic inaccuracies of the MLP algorithm. In this work the analytical simulator is used to explore and firstly compare the imaging performances of list-mode and integration-mode detector configurations with proton pencil beam scanning. RESULTS: The intrinsic inaccuracies of the MLP algorithm affect the imaging performances of list-mode detector configuration, which nevertheless remains superior to integration-mode detector configuration for iCTs. For relatively higher proton statistics, comparable or better imaging performances are offered by integration-mode detector configuration for iRads (upto 29.2% of WET difference). Uncertainties of proton trajectories due to beam spot size are shown to compromise the imaging performances of integration-mode detector configuration, but also to affect the accuracy of the MLP algorithm for list-mode detector configuration. CONCLUSIONS: Based on MCS model in non-uniform water equivalent materials, the proposed simulation environment can serve for development and testing of dedicated imaging methodologies prior to and in combination with realistic Monte Carlo simulations.

Comparative Monte Carlo study on the performance of integration- and list-mode detector configurations for carbon ion computed tomography.

Ion beam therapy offers the possibility of a highly conformal tumor-dose distribution; however, this technique is extremely sensitive to inaccuracies in the treatment procedures. Ambiguities in the conversion of Hounsfield units of the treatment planning x-ray CT to relative stopping power (RSP) can cause uncertainties in the estimated ion range of up to several millimeters. Ion CT (iCT) represents a favorable solution allowing to directly assess the RSP. In this simulation study we investigate the performance of the integration-mode configuration for carbon iCT, in comparison with a single-particle approach under the same set-up. The experimental detector consists of a stack of 61 air-filled parallel-plate ionization chambers, interleaved with 3 mm thick PMMA absorbers. By means of Monte Carlo simulations, this design was applied to acquire iCTs of phantoms of tissue-equivalent materials. An optimization of the acquisition parameters was performed to reduce the dose exposure, and the implications of a reduced absorber thickness were assessed. In order to overcome limitations of integration-mode detection in the presence of lateral tissue heterogeneities a dedicated post-processing method using a linear decomposition of the detector signal was developed and its performance was compared to the list-mode acquisition. For the current set-up, the phantom dose could be reduced to below 30 mGy with only minor image quality degradation. By using the decomposition method a correct identification of the components and a RSP accuracy improvement of around 2.0% was obtained. The comparison of integration- and list-mode indicated a slightly better image quality of the latter, with an average median RSP error below 1.8% and 1.0%, respectively. With a decreased absorber thickness a reduced RSP error was observed. Overall, these findings support the potential of iCT for low dose RSP estimation, showing that integration-mode detectors with dedicated post-processing strategies can provide a RSP accuracy comparable to list-mode configurations.