Purpose: Firstly, to develop fastCAT, a cone beam CT (CBCT) simulation platform capable of modeling the physics, geometries, and phantoms of commercial linacs with high computational efficiency. Secondly, the platform was used to conduct a simulation study investigating novel MV imaging beam and detector combinations.
Methods: FastCAT combines GPU ray-tracing with Monte Carlo (MC) methods for improved simulation efficiency. FastCAT requires photon spectra and detector optical spread functions (OSFs) as input. 2.5 and 6 MV photon spectra were modelled in EGSnrc using carbon, aluminum, and tungsten targets. kV x-ray tube spectra were simulated analytically. OSFs were simulated in Topas for cadmium tungstate (CWO), gadolinium oxysulfide (GOS), and cesium iodide (CsI) detectors. FastCAT contrast and noise magnitude were validated using a Truebeam linac and a Catphan 504 phantom for 100 kVp and 6 MV CBCTs, respectively. The simulation study calculated CNRs for all detector/beam combinations using a contrast phantom containing rib and spongiosa bone, lung, and adipose tissue inserts. Detector modulation transfer function (MTF) for each beam/detector combination was calculated using an angled slit phantom.
Results: Truebeam contrast and noise magnitude matched all fastCAT values within the 99% confidence interval. FastCAT simulation times were 51s and 77s for the kV and MV CBCT, respectively, significantly shorter than the estimated 1-year long full MC simulations.The simulation study showed CNRs for the carbon 2.5 MV/CWO combination were more than 1.5 times greater than the next best 120 kVp/CsI combination, while MTF 50 values were only 1.2% lower for the carbon 2.5 MV/CWO combination.
Conclusion: Our finding makes a compelling case that clinical CBCT imaging could be greatly improved using carbon target beams with a CWO detector.