In Silico Phase-Contrast X-Ray Imaging of Anthropomorphic Voxel-Based Phantoms

Abstract

Propagation-based phase-contrast X-ray imaging is an emerging technique that can improve dose efficiency in clinical imaging. In silico tools are key to understanding the fundamental imaging mechanisms and develop new applications. Here, due to the coherent nature of the phase-contrast effects, tools based on wave propagation (WP) are preferred over Monte Carlo (MC) based methods. WP simulations require very high wave-front sampling which typically limits simulations to small idealized objects. Virtual anthropomorphic voxel-based phantoms are typically provided with a resolution lower than imposed sampling requirements and, thus, cannot be directly translated for use in WP simulations. In the present paper we propose a general strategy to enable the use of these phantoms for WP simulations. The strategy is based on upsampling in the 3D domain followed by projection resulting in highresolution maps of the projected thickness for each phantom material. These maps can then be efficiently used for simulations of Fresnel diffraction to generate in silico phase-contrast X-ray images. We demonstrate the strategy on an anthropomorphic breast phantom to simulate propagation-based phase-contrast mammography using a laboratory micro-focus X-ray source.

Phase-contrast virtual chest radiography

“…The model was upsampled to ∼3-8-μm resolution corresponding to z = 2 to 12 m according to ref. ( 18 ). Preprocessing was performed in MATLAB on a computer node with 1 TB RAM and 4×12 CPU cores Intel E7-8857v2 Ivy Bridge.…”

“…Fig. 1 ) of the processed virtual patient using a computational wave-propagation approach ( 18 ). Three different settings were simulated: Phase-contrast chest radiography, simulated with a monochromatic beam (60 keV) at z = 2 to 12 m. Conventional chest radiography, simulated with a 120-kVp tungsten spectrum (with 4.3-mm aluminum filtration ( 42 )) at z = 0 m. Control radiography, simulated with a monochromatic beam (60 keV) at z = 0 m. …”

“…Next, we performed a computationally expensive edge-preserving 3D upsampling of the virtual patients (reducing voxel sizes from 100 μm to sub-10 μm) followed by compression into tissue thickness maps sufficient for generating virtual chest radiographs with conventional or phase-contrast settings. The latter is modeled by numerically evaluating the Fresnel-diffraction integral ( 18 ). Lastly, we applied postprocessing on the radiographs using a clinically used commercial software application to further enhance the realism of the virtual radiographs.…”

“…Computational modeling of clinical phase-contrast X-ray imaging is particularly challenging due to the typical few-μm wavefront-sampling requirements for modeling the interference effects in conjunction with the large sizes of human body parts (∼tens of centimeters). To this end, we recently demonstrated a preprocessing strategy unlocking PBI simulations on realistic virtual models of human anatomy ( 18 ).…”

X-ray differential phase-contrast imaging simulations with Geant4

“…Various tools to simulate phase-contrast effects have been developed over the years and they can be grouped into three categories, namely wave propagation (WP), Monte Carlo (MC), and hybrid simulations. In WP simulations, x-rays are treated as waves and phase effects are usually computed following the Fresnel propagator formalism: interesting WP implementations can be found in the field of PBI [14,15] and GI [16]. Conversely, MC implementations intrinsically follow the ray-tracing approach.…”