PurposeTo establish a dose volume–effect relationship for predicting late rectal complication (LRC) in locally advanced cervical cancer patients treated with external beam radiotherapy (EBRT) followed by combined intracavitary/interstitial brachytherapy (IC/IS-BT).Materials and MethodsA retrospective analysis was performed in 110 patients with locally advanced cervical cancer who underwent definitive radiotherapy combined with IC/IS-BT from July 2010 to September 2018. We report the 90% of the target volume receiving the minimum dose for high risk clinical target volume (HR-CTV D90) and intermediate risk clinical target volume (IR-CTV D90), and the minimum doses to the most exposed 0.1, 1, and 2 cm³ D0.1cm3,D1cm3,D2cm3 doses at the International Commission on Radiation Units and Measurements (DICRU) for organs at risk (OARs). The total dose of EBRT plus brachytherapy was transformed to the biologically equivalent dose in 2 Gy fractions (EQD2) with α/β value of 10 Gy for target, 3 Gy for organs at risk using the linear quadratic model. The morbidity was scored according to the Radiation Therapy Oncology Group (RTOG) criteria. The Probit model was used to establish a prediction model on rectum between the organs at risk for dose and LRC. The receiver operating characteristic (ROC) curve was used to evaluate the predictive value of dose volume parameters for LRC.ResultsThe median follow-up time was 72.3 months. The mean ( ± standard deviation) D2cm3,D1cm3,D0.1cm3, and DICRU values of rectum were 64.72 ± 7.47 GyEQD2, 70.18 ± 5.92 GyEQD2, 79.32 ± 7.86 GyEQD2, and 67.22 ± 7.87 GyEQD2, respectively. The Probit model showed significant relationships between D1cm3 or D0.1cm3, and the probability of grade1–4, grade 2–4 rectal events at 1 year, and between D1cm3and the probability of grade2–4 rectal events at 3 and 5 years. The dose values for 10% complication rates (ED10) of D1cm3were 74.18 (70.42–76.71) GyEQD2, 67.80 (59.91, 71.08) GyEQD2, 66.37 (52.00, 70.27) GyEQD2 for grade 2–4 with rectal morbidity at 1, 3, and 5 years, respectively.ConclusionOur study proved that D1cm3 andD0.1cm3were considered as useful dosimetric parameters for predicting the risk of grade1–4 and grade2–4 LRC at 1-year, and D1cm3might be an indicator for predicting grade2-4 LRC at 3/5years. The patients with rectal D1cm3>66.37–74.18 GyEQD2 should be closely observed for grade2–4 LRC.
Development and validation of a deep learning algorithm for auto-delineation of clinical target volume and organs at risk in cervical cancer radiotherapy