A novel reconstruction method for 3D cone beam CT images in proton therapy using RTK

Abstract

Purpose: CBCT reconstruction research requires access to different algorithms. Offline reconstruction for the Varian ProBeam CBCT system has not been reported. We aim to: 1) develop a novel offline reconstruction method for ProBeam CBCT and 2) simulate ultra-low-dose CBCT images. Methods: Straightforward reconstruction of ProBeam projection images with RTK results in significant artefacts. Our reconstruction pipeline includes processing on projection images as well as the reconstructed CBCT. 10 air calibration images are used to correct for bowtie filter, source, and detector shifts during gantry rotation (flood field flex). Beam hardening by the bowtie filter is corrected using an empirically-derived attenuation space map. RTK version 1.4 is used for filtered backprojection and slice averaging is then applied. The pipeline was validated by comparison to clinically-reconstructed CBCT images.To test the simulation of ultra-low-dose images, a Catphan 503 phantom was scanned at 9 mAs exposures (35-811 mAs/projection, 100 kV, SDD 370 cm). The signal-noise relationship was measured from phantom projection data to inform the amount of non-uniform Gaussian noise to be added to simulate low dose scans. Ultra-low-dose CBCTs at 30% and 10% exposure were simulated using datasets from paediatric patients treated on ProBeam. Results: The exposure and hardening corrections completely eliminate artefacts. The processed reconstructed images show excellent visual agreement with clinically reconstructed CBCT images. SNR measured in simulated phantom projection data shows good agreement with that in acquired images. Simulated low-exposure scans of paediatric patients are noisier but maintain good visibility of bone and some soft tissue, even at 10% exposure. Conclusion: We successfully implemented offline reconstruction for Varian Probeam CBCT images. Correcting for flood field image flex is essential in large-geometry imaging systems. Our methods allow optimisation of CBCT acquisition and reconstruction, especially relevant for paediatric treatments, and can be easily applied to other scan types e.g. 4DCBCT.