Affiliations: [a] Département de physique, de génie physique et d’optique, Université Laval, Québec (Québec), Canada
| [b] Centre de Recherche sur le Cancer, Université Laval, Québec (Québec), Canada
| [c] Département de Radio-Oncologie and Centre de Recherche du CHU de Québec, Québec (Québec), Canada
Abstract: BACKGROUND:Four-dimensional cone beam computed tomography allows for temporally resolved imaging with useful applications in radiotherapy, but raises particular challenges in terms of image quality and computation time. OBJECTIVE:The purpose of this work is to develop a fast and accurate 4D algorithm by adapting a GPU-accelerated ordered subsets convex algorithm (OSC), combined with the total variation minimization regularization technique (TV). METHODS:Different initialization schemes were studied to adapt the OSC-TV algorithm to 4D reconstruction: each respiratory phase was initialized either with a 3D reconstruction or a blank image. Reconstruction algorithms were tested on a dynamic numerical phantom and on a clinical dataset. 4D iterations were implemented for a cluster of 8 GPUs. RESULTS:All developed methods allowed for an adequate visualization of the respiratory movement and compared favorably to the McKinnon-Bates and adaptive steepest descent projection onto convex sets algorithms, while the 4D reconstructions initialized from a prior 3D reconstruction led to better overall image quality. CONCLUSION:The most suitable adaptation of OSC-TV to 4D CBCT was found to be a combination of a prior FDK reconstruction and a 4D OSC-TV reconstruction with a reconstruction time of 4.5 minutes. This relatively short reconstruction time could facilitate a clinical use.
