Program Information
A New Approach to Realize and Reconstruct Multi-Energy Cone-Beam CT for Radiotherapy
B Li1,2*, C Shen1 , Y Chi1 , M Yang1 , X Jia1 , (1) University of Texas Southwestern Medical Center, Dallas, TX, USA (2) Southern Medical University, Guangzhou, Guangdong, China
Presentations
WE-DE-605-1 (Wednesday, August 2, 2017) 10:15 AM - 12:15 PM Room: 605
Purpose: Multi-energy computed tomography has a number of potential applications in diagnosis and therapy. However, high system cost obstructs its step into routine clinical practice. In this study, we propose to realize multi-energy cone beam CT (ME-CBCT) on a conventional CBCT system that is widely used for radiotherapy image guidance. To achieve high-quality images, we also propose an iterative reconstruction algorithm.
Methods: A kVp switching technique was realized using the CBCT platform on a TrueBeam linear accelerator via its developer mode. It acquired projections by switching kVp levels alternatively during gantry rotation. For this kVp-switching based ME-CBCT acquisition, x-ray projections of each energy channel are only a subset of all the acquired projections. We proposed a spatial spectral non-local means (ssNLM) method to iteratively reconstruct high-quality ME-CBCT images incorporating inherent non-local correlations between image patches along both spatial and spectral directions, hence suppressed noisy and streak artifacts induced from undersampled data. GPU implementation was used to achieve a high computational efficiency. We comprehensively validated our method in both simulation and experimental studies. In simulation, we generated projection data of a numerical NCAT phantom by using Monte-Carlo simulation. In experiments, we tested the method with a Catphan, a Gammex and an anthropomorphic head phantoms.
Results: We compared three reconstruction methods: FDK, conventional NLM and the proposed ssNLM. For both simulation and experimental studies, visual inspection found that images of FDK and NLM methods existed noise and streak artifacts, while ssNLM method effectively removed them while preserving fine structures. Quantitatively, ssNLM had the best SSIM, FSIM and RMSE values. CNR measurement demonstrated that ssNLM were more than 3 times higher than FDK and around 50% higher than conventional NLM method.
Conclusion: A comprehensive set of simulation and experimental studies demonstrated the feasibility of our ME-CBCT scheme and capability of achieving high-quality images.
Contact Email: