Abstract
BACKGROUND: In online adaptive proton therapy (OAPT), additional treatment verification techniques are crucial for detecting treatment deviations and acting as a safety net, as phantom-based patient-specific quality assurance methods are not applicable. Prompt gamma imaging (PGI) has the potential for online treatment verification without adding dose nor prolonging treatment. PGI has proven to detect relevant anatomical changes under real-world clinical conditions. Range probing (RP) has shown its clinical applicability for in-vivo proton range assessment before treatment, thereby helping to address uncertainties. It has been proven useful for quality control of cone-beam CT-based synthetic CTs, suggesting its adoption into OAPT. PURPOSE: The performance of PGI and RP, two different yet complementary treatment verification methods, was compared by consecutive measurements within one experimental setup. Anatomical changes (AC) and setup errors (SE) were mimicked in an anthropomorphic head phantom. METHODS: The PGI-system was positioned beneath the phantom while the RP-system was positioned distally, allowing a simultaneous setup of both systems at a horizontal gantry angle. A brain target was irradiated with a 1-field pencil-beam scanning treatment plan and a low-dose RP-plan with high-energy protons that passed through the phantom. Upstream and downstream positioned water-equivalent material slabs of 2/3/5 mm or 5 mm thicknesses mimicked AC within the beam path or beyond the target, respectively. Additionally, the couch was shifted 2 and 3 mm in left, down and upstream direction in beam's eye view (BEV), mimicking SE. Both plans were delivered and monitored 10 times for each AC and SE configuration. Geometrical range shifts measured with PGI were converted to water-equivalent thickness range shifts for direct comparison between PGI and RP results. RESULTS: Both systems detected range deviations relevant for the treatment field, caused by AC within the clinical beam path. RP measurements were more precise for all scenarios (RP: 1σ ≤ 0.3 mm, PGI: 1σ ≤ 0.8 mm). PGI was similarly accurate for the 2 mm slab while slightly underestimating 3 and 5 mm slabs (≤ 0.7 mm). Only RP detected AC beyond the target, which are irrelevant for the monitored treatment field. Both systems detected expected range shifts of all SEs. RP was more accurate than PGI (RP: within 0.3 mm, PGI: within 0.8 mm). The couch movement left in BEV was detected with slightly higher precision using PGI (RP: 1σ ≤ 1.0 mm, PGI: 1σ ≤ 0.9 mm), while the precision of detecting the couch movement down in BEV was the same for both systems (1σ ≤ 0.7 mm). Only PGI recognized couch movements upstream (accuracy within 0.4 mm). CONCLUSION: Both PGI and RP precisely detected introduced AC relevant for the monitored treatment field. SE were detected, but with greater uncertainty. While the current implementation of RP enables pretreatment range verification after setup imaging with a low-dose RP-field, PGI enables treatment verification during field delivery, detecting range deviations relevant for the treatment field. The simultaneous setup of PGI and RP clearly demonstrated their compatibility in a clinical setting, the unique advantages of each system and their crucial role as safety nets in OAPT.