Methodological Study of Geometric Deformation for CBCT in Proton Therapy Gantry

Rosa Seabra, José Carlos [UCL] Brousmiche, Sébastien [UCL] Labarbe, Rudi [IBA] Lee, John Aldo [UCL] Macq, Benoît [UCL]

Cone-Beam CT systems allow in-situ imaging for guiding radiation therapy by 3D reconstruction of radiographs acquired during gantry rotation. However, deformation of gantry geometry is a primary source of reconstruction artifacts including edge blurring, loss of spatial resolution, truncation, and streaks [1]. Deviations from the ideal X-ray source and detector trajectories are caused by reproducible and random deformation. Whilst the first can be corrected by an appropriate calibration procedure, the second type shows an unpredictable pattern although its magnitude is small compared to reproducible deformation. Several methods have been proposed for correcting geometric non-idealities in tomographic imaging systems such as C-arm CT, cone-beam breast CT or LINAC-integrated cone-beam CT [2,3,4]. However, the magnitude of geometric deformation in a proton therapy gantry is typically higher than in other systems due to gantry geometry with larger source to detector distance, larger bearing structure and gravity-induced flex in the support arms. Hence, these differences limit the application of calibration methods described in the literature and justify the need for a complete simulation framework on geometric deformation in a proton therapy gantry. This study is part of the development of a calibration routine for a proton therapy gantry using a phantom with radio-opaque markers (spheres) and full sets of radiographs acquired while rotating the gantry. The calibration parameter values comprising source displacement, flat panel displacement and rotation angles, and gantry angle offset are found by using projection matrices of the object onto the detector. The method implies for each gantry angle the projection of phantom markers onto the flat panel assuming no deformation, the identification of spheres on the radiograph, and registration of nominal and measured markers positions using optimization techniques. This paper aims at developing a comprehensive test bench for assessing the geometric deformation of a proton therapy gantry, comprising a software simulator validated by in-situ measurements. A set of simulations is performed to quantify the impact of phantom geometry, marker detection rate, and incorrect estimation of marker position and choice of optimization technique on the accuracy of the calibration routine. The correlation between source and detector offsets, which is known to make the optimization unstable, is also investigated. Moreover, this test bench provides an effective tool for evaluating the performance of the calibration routine and straightforward assessment of potential improvements to the current technique. Finally, the knowledge acquired with the proposed methodological study is used to set optimal conditions for the application of the calibration routine in data sets acquired with a proton therapy gantry. Preliminary measurements of gantry geometry enable to fit models describing the gantry deformation depending on the gantry angle. These deformation models have been used to improve reconstruction in Cone-Beam CT.