ESTRO 2022 - Abstract Book

S1374

Abstract book

ESTRO 2022

In this study, we demonstrate a novel system design for proton radiography and use two imaging approaches to generate proton radiographs. The novel system design consists of a monolithic scintillator detector which captures complete proton beam deposition and circumvents the need to modulate proton beams. Proton radiographs are generated using projections from 2 cameras: (a) by integrating light along the beam axis (beam-integration method), and (b) by recording changes to the proton Bragg peak (BP) location for the beam as it travels through the phantom (percentage depth light or PDL method). Materials and Methods The imaging system consists of a monolithic plastic scintillator detector and two CCD cameras for imaging the scintillation light distribution (refer Figure 1). A 45° angled mirror redirects light to camera 2 without directly exposing it to ionizing radiation. Camera 2 generates images by integrating light along the beam axis. The light intensity is converted into water equivalent thickness (WET) by plotting an energy-specific calibration graph. A polynomial fit to this graph is used to calculate phantom WETs. Camera 1 images BP locations of scanned pencil beams. The radiograph is reconstructed by comparing the PDL profiles for these beams that pass through the phantom with pristine PDL profiles (without phantom) shifted by a convolution of the WET. A curvelet minimization method is applied to improve image resolution. Gammex phantoms (solid water, cortical bone) were imaged using the system (refer Figure 1). The relative percentage accuracy, (WET expt – WET calc ) / WET calc * 100, was used to evaluate the system’s performance in retrieving WET values for phantoms.

Results The system characterization studies included assessing its linearity (R 2 = 1) over two orders of magnitude change in proton dose, camera resolution (0.44 mm/pixel), and short (0.37%) and medium term (2% over 14 weeks) stability. Figure 2 shows proton radiographs for cortical bone generated using the two reconstruction methods. The resolution for the beam- integration method is limited by the CCD sensor pixel size. Blurring at edges is due to proton scatter occurring within the phantom as well as in the scintillator volume. The spatial resolution of the PDL method is limited by the pencil beam size and the image therefore looks pixelated. Relative percentage accuracies of − 0.18 ± 0.35% and − 2.94 ± 1.20% for solid water and cortical bone, respectively, were obtained from the beam-integration method, while accuracies of − 0.29 ± 3.11% and − 0.75 ± 6.11% for solid water and cortical bone were calculated for the PDL method.