Proton Therapy Treatment Simulations with the Bergen DTC Prototype for Range Verification

This dataset consists of 36258 proton therapy Monte Carlo simulations using GATE (Jan et al. 2004) version 9.2 and Geant4 (Agostinelli et al. 2003; Allison et al. 2006; Allison et al. 2016) version 11.0.0. The treatment target is the pediatric head phantom 715-HN by CIRS Inc. (Norfolk, VA, United States), digitized by Giacometti et al. (2017). Distal to the phantom, is a simplified geometry of the digital tracking calorimeter designed by the Bergen pCT collaboration for proton computed tomography (Alme et al. 2020). Each simulation is a single pencil beam with \(10^7\) primary protons. To determine valid spots to use, probing simulations with \(10^5\) primaries were conducted, where the beam position, energy, and phantom rotation were varied: 10 mm interval in the xy-plane for the beam position, 3 mm water range interval for the beam energy, and 30° rotations of the phantom. If no primary was found in the detecter after the probing simulation, it is a valid spot for the dataset. Resulting hit files, containing simulated detector readout, can be found in treatment_simulations.tar.gz. Each simulation is accompanied by a metafile in JSON format, specifying all parameters used to produce the output. E.g., beam position and energy can be found in "parameters" -> "beam_spot_x", "beam_spot_y", "beam_energy" and the phantom rotation can be found in "parameters" -> "phantom_rotation_angle". Simulation data has been pre-processed with two steps. First, hits for the same track in the same layer are binned together into a single entry by averaging the positions (posX, posY, posZ) and summing the energy depositions (edep). Second, the energy deposition values are discretized into a cluster size and back into MeV, to represent the energy resolution of the ALPIDE chip (Alme et al. 2020). Clusters of size 0 are removed. Additionally, some of the unused GATE output columns are removed and a column specifying the detector layer is created through the formula layer = 2*volumeID[2] + volumeID[3]. Details about the meaning of the other output columns can be found in the GATE documentation.