Dosimetric Evaluation of CBCT-Guided Setup errors and Adaptive Radiotherapy Replanning in Proton Therapy for Lung Cancer

Abstract

Abstract Background To investigate the dosimetric impact of setup errors in cone-beam computed tomography (CBCT)-guided proton therapy for lung cancer and to evaluate the role of adaptive radiotherapy (ART) based on a sequential adaptive planning approach for maintaining optimal dose distribution after reductions in target volume. Methods Patients with lung cancer who underwent proton radiotherapy at our institution between April 2024 and January 2026 were retrospectively evaluated. The first cohort (n = 29) received treatment according to the original plan throughout the entire course. CBCT images acquired during each treatment session were used to calculate mean setup deviations in the X, Y, and Z directions. These average shifts were applied to generate simulated treatment plans, and dosimetric parameters for the CTV and OARs were compared with those of the original plans. In the second group of patients (n = 17), one adaptive radiotherapy intervention was implemented during treatment. The initial treatment plan was termed Plan00. Following the detection of significant target shrinkage on CBCT, a repeat CT scan was acquired. The dose delivered before replanning was recorded as Plan01, while a revised plan (Plan1) was developed using the new CT dataset. After dose registration and accumulation of Plan01 and Plan1, the resulting dose-volume metrics were compared with those of Plan00 to evaluate the dosimetric impact of adaptive replanning relative to a non-adaptive treatment strategy. Changes in dosimetric parameters were analyzed, including lung V 5 , V 20 and D mean , spinal cord D 1cc and D max , esophageal V5, and D mean . Results Average setup errors along the X, Y, and Z axes were 0.08 ± 2.71 mm, 0.53 ± 0.36 mm, and 0.23 ± 1.03 mm, respectively, with maximum absolute deviations of 14 mm, 29 mm, and 14 mm. The greatest positioning variability was observed in the Y direction, where 7% of errors exceeded 5 mm. Errors ≤ 3 mm occurred in 81.5%, 76.6%, and 93.3% of fractions, whereas errors ≤ 5 mm were observed in 99.0%, 92.8%, and 94.5% of fractions for the X, Y, and Z axes, respectively. Compared with the simulated plan, the original plan achieved 104.31% ± 14.41%, 102.23% ± 11.40%, 98.84% ± 9.37%, and 100.26% ± 9.66% of the corresponding CTV D 98% , D 95% , D 2%, and D mean values, respectively. Ratios for spinal cord D 1cc , spinal cord D max , lung V 5 , lung V 20 , bilateral lung D mean , esophageal V 5 , and heart D mean were 90.60% ± 24.05%, 87.23% ± 21.57%, 91.11% ± 45.12%, 101.01% ± 36.17%, 100.22 ± 14.46%, 105.50% ± 20.36%, and 102.15% ± 22.05%, respectively. The simulated plan showed significantly lower CTV D9 8%, D 95% , and D mean values than the original plan (all P 3 on the initial planning CT to 97 ± 69 cm 3 on the repeat CT, representing a mean reduction of 35 ± 55 cm 3 . This reduction was statistically significant (t = 2.54, P 20 (20.66 ± 8.14% versus 17.47 ± 7.78%), and whole-lung D mean (6.26 ± 2.98 GyE versus 5.33 ± 2.78 GyE) (all P 30 decreased from 8.36 ± 14.84% to 4.22 ± 3.89%, and spinal cord D max decreased from 23.04 ± 15.61 GyE to 21.55 ± 13.59 GyE; however, neither difference was statistically significant (P > 0.05). Conclusion Proton therapy is highly susceptible to setup uncertainties and anatomical variations during treatment. CBCT-guided image verification plays a critical role in maintaining treatment accuracy for lung cancer. Adaptive replanning based on imaging-detected changes may help preserve target dose coverage while further reducing radiation exposure to OARs.