Monte Carlo model of a prototype flat-panel detector for multi-energy applications in radiotherapy
Ikechi Ozoemelam,Marios Myronakis,Thomas C Harris,Pablo Corral Arroyo,Pascal Huber,Matthew W Jacobson,Yue-Houng Hu,Rony Fueglistaller,Mathias Lehmann,Daniel Morf,Ross I Berbeco
DOI: https://doi.org/10.1002/mp.16689
Abstract:Background: The incorporation of multi-energy capabilities into radiotherapy flat-panel detectors offers advantages including enhanced soft tissue visualization by reduction of signal from overlapping anatomy such as bone in 2D image projections; creation of virtual monoenergetic images for 3D contrast enhancement, metal artefact reduction and direct acquisition of relative electron density. A novel dual-layer on-board imager offering dual energy processing capabilities is being designed. As opposed to other dual-energy implementation techniques which require separate acquisition with two different x-ray spectra, the dual-layer detector design enables simultaneous acquisition of high and low energy images with a single exposure. A computational framework is required to optimize the design parameters and evaluate detector performance for specific clinical applications. Purpose: In this study, we report on the development of a Monte Carlo (MC) model of the imager including model validation. Methods: The stack-up of the dual-layer imager (DLI) was implemented in GEANT4 Application for Tomographic Emission (GATE). The DLI model has an active area of 43×43 cm2 , with top and bottom Cesium Iodide (CsI) scintillators of 600 and 800 μm thickness, respectively. Measurement of spatial resolution and imaging of dedicated multi-material dual-energy (DE) phantoms were used to validate the model. The modulation transfer function (MTF) of the detector was calculated for a 120 kVp x-ray spectrum using a 0.5 mm thick tantalum edge rotated by 2.5o . For imaging validation, the DE phantom was imaged using a 140 kVp x-ray spectrum. For both validation simulations, corresponding measurements were done using an initial prototype of the imager. Agreement between simulations and measurement was assessed using normalized root mean square error (NRMSE) and 1D profile difference for the MTF and phantom images respectively. Further comparison between measurement and simulation was made using virtual monoenergetic images (VMIs) generated from basis material images derived using precomputed look-up tables. Results: The MTF of the bottom layer of the dual-layer model shows values decreasing more quickly with spatial frequency, compared to the top layer, due to the thicker bottom scintillator thickness and scatter from the top layer. A comparison with measurement shows NRMSE of 0.013 and 0.015 as well as identical MTF50 of 0.8 mm1 and 1.0 mm1 for the top and bottom layer respectively. For the DE imaging of the DE-phantom, although a maximum deviation of 3.3% is observed for the 10 mm aluminum and Teflon inserts at the top layer, the agreement for all other inserts is less than 2.2% of the measured value at both layers. Material decomposition of simulated scatter-free DE images gives an average accuracy in PMMA and aluminum composition of 4.9% and 10.3% for 11-30 mm PMMA and 1-10 mm aluminum objects respectively. A comparison of decomposed values using scatter containing measured and simulated DE images shows good agreement within statistical uncertainty. Conclusion: Validation using both MTF and phantom imaging shows good agreement between simulation and measurements. With the present configuration of the digital prototype, the model can generate material decomposed images and virtual monoenergetic images.