Abstract
BACKGROUND: Accurate target delineation is essential for precise delivery of proton therapy. A magnetic resonance imaging (MRI)-only radiotherapy workflow may improve delineations, and thus improve proton therapy, but requires reliable synthetic computed tomography (sCT) for accurate dose calculations. PURPOSE: The aim was to evaluate the dosimetric accuracy of commercial software-based sCT for proton dose calculation for head and neck cancer (HNC), and benchmark it against the accuracy of photon dose calculation on the same sCT. METHODS: MRI and planning CT (pCT) were acquired for 20 HNC patients before receiving photon therapy. A commercial software created MRI-based sCTs, which were co-registered to pCTs. For each patient, photon and proton plans based on pCT, prescribing 68 Gy to the target, were created in RayStation and recalculated on sCT. Dose-volume histogram (DVH) metrics and gamma index (2%/2 mm criteria) were analyzed, comparing pCT- and sCT-calculated dose. Local and global gamma index were calculated for low and high-dose thresholds. A gamma pass rate (GPR) was calculated for each plan and gamma index type. RESULTS: For photon and proton plans, the median difference in DVH metrics between pCT- and sCT-calculated dose was < 0.4 Gy with an interquartile range < ± 0.7 Gy for all structures and metrics, except for mean dose to oesophagus (median = -0.3 Gy, range: -2.0 Gy to 0.7 Gy), oral cavity (median = 1.8 Gy, range: 0.6 Gy to 3.0 Gy), and larynx (median = -0.1 Gy, range: -1.5 Gy to 0.6 Gy) for proton plans. The median GPR for photon plans was > 97.7%, while for proton plans it was > 93.3% except for body GPR with low-dose threshold. CONCLUSION: While sCT seems feasible for photon therapy, poorer agreement between pCT- and sCT-calculated proton dose was found. Patient positioning differences between CT and MRI may partly explain the discrepancies between pCT- and sCT-calculated proton dose.