Offline adaptive radiation therapy for prostate cancer: using daily CBCT and deformable image fusion for correct replanning

Abstract

INTRODUCTION Adaptive radiation therapy (ART) is an established clinical practice, especially for treatments requiring rapid changes due to organs-at-risk (OAR) that might influence the target position. Adapting the procedure to a case-to-case basis involves combining different tools, such as scanning pretreatment images, clinically assessing the need for adaptation, replanning a new treatment, and guaranteeing the final quality of the entire process. Modern radiation therapy equipment enables multiple optimization strategies, both online and offline. The primary aim of this study is to define an offline ART procedure to correct the replanning of prostate treatments according to objective evaluation criteria. MATERIALS AND METHOS The simulation and treatment protocols for prostate patients involve emptying the rectum when needed and ensuring that the bladder is filled with adequate urine volume. To comply with the simulation conditions during the treatment, daily cone-beam computed tomography (CBCT) images are acquired and controlled on a daily basis. The image-guided radiation therapy (IGRT) protocol provides a rigid fusion of the images acquired in the bunker with those collected from the simulation CT. For this study, we selected 23 patients with prostate adenocarcinoma (medium and low risk) treated with 40 fractions, with a daily dose of 2 Gy (80 Gy) at UPMC San Pietro FBF Advanced Radiotherapy Center in Rome, from October 2018 to May 2019. During the treatment, patients were placed in the supine position, with their arms on their chest and legs restrained by an immobilization device (ProSTEP™ Klarity). The offline ART workflow required pretreatment verifications, registration with the simulation images, and calculation of the rectum and bladder filling variations. The analysis was performed using the Velocity v4.0 software (Varian Medical System, Palo Alto CA). At the end of the Velocity-based software-automated process, the CT and CBCT images were used to generate an aCT (adaptive CT). Organs of interest were contoured on the aCT automatically. The Dice coefficient and the dispersion and distribution statistical indexes were taken into consideration to ensure accurate qualitative comparison. RESULTS Percentage dispersion of the rectum volume values was higher in Group A. Distribution of rectum volume variation percentage in Group A had an IQR = 5,55% (Q1=-4,06%; Q2= -1,13%; Q3= 1,49%), whereas Group B had an IQR= 4,24% (Q1= -2,50%; Q2= 2,09%; Q3= 1,75%). Percentage dispersion of the bladder volume values was higher in Group A. Distribution of bladder volume variation percentage in Group A had an IQR = 9,65% (Q1=-7,34%; Q2= -2,32%; Q3= 2,31%), whereas Group B had an IQR = 12,13% (Q1= -7,18%; Q2= -1,56%; Q3= 4,96%). The Dice coefficient in Group A showed an average daily superimposition of the bladder of 0,91 ± 0,07, whereas in Group B this was 0,87 ± 0,10. In both groups, the rectum volume had an average Dice coefficient of 0,89 ± 0,09. CONCLUSIONS The results show that the Dice coefficient can be useful to establish whether the volume localization can be superimposed to the simulation CT. Based on our practice, we suggest that the offline ART protocol should be verified over the first five therapy fractions, representing an adequate window to assess the need for replanning. Because this index does not consider the volumes but only the possibility of their geometric superimposition, we recommend checking the mean OAR volumes when using an offline ART workflow. This is particularly important for the bladder, which is more susceptible to this kind of change than variations in its localization.