Session Item

As for external beam radiotherapy (EBRT), also for brachytherapy (BT) treatment planning is a time-consuming task. Currently available commercial BT treatment planning systems (TPS) require interactive trial-and-error dwell time optimization to obtain an acceptable dose distribution, dependent on planner experience and allotted time. We built a fully automated treatment planning workflow for adaptive fractionated HDR BT for locally advanced cervical cancer after EBRT, in line with EMBRACE II protocol planning aims and objectives. The system was validated by comparing automated plans with manually generated clinical plans. 

Our in-house TPS for automated multi-criterial plan generation for EBRT was extended for HDR-BT. A process was developed to automatically establish daily adaptive BT planning, based on delivered EBRT dose and BT doses of previous fractions, following the EMBRACE II protocol and results in clinically deliverable treatment plans. For 15 previously treated cervical cancer patients, the clinical scans with contours and dwell positions of all 3 or 4 BT applications were used and BT dose distributions were automatically generated, with a high risk CTV (CTVHR) coverage equal to that in the clinical plan. Apart from requirements for targets and OARs, there were also non-dosimetric objective functions to prevent too high dwell time gradients and unbalanced intracavitary, ovoid and interstitial needle usage.  Fully automated adaptive BT treatment planning was compared with clinical planning using EQD2Gy for the complete treatment (EBRT+BT). Moreover, the sum of all dwell times (Efficiency), and the ratio of dwell times outside CTVHR/sum of all dwell times (Outside/Total ratio) were evaluated. Wilcoxon signed ranked test was used for statistical evaluation.

Autoplanning resulted in lower OAR EQD2Gy compared to clinical planning for similar CTVHR coverage (Figure 1). Intermediate risk CTV (CTVIR) coverage was lower with automated plans compared to clinical plans, but always within protocol constraints. D₂­_cc for bladder, rectum, sigmoid and bowel were in favour of autoplanning with average gains of 5.1 Gy [1.5 - 10.5 Gy, p<0.001] , 2.1 Gy [-1.5 - 7.0 Gy, p=0.002] , 2.4 Gy [-1.2 - 5.8 Gy, p<0.001] and 2.6 Gy [0.2 - 7.6 Gy, p<0.001], respectively. Moreover, large gains were also observed for the rectovaginal point (3.7 Gy [-2.3 - 10.5 Gy, p<0.001]) and D_max in the vaginal tops left (11.7 Gy [-12.2 - 38.2 Gy, p=0.005]) and right (13.9 Gy [-4.7 - 42.2 Gy, p<0.001]). Autoplanning reduced the Efficiency on average by 13% with an Outside/Total ratio 33% lower compared to the clinical plans. Optimization times for autoplanning were in the order of 3-6 minutes per plan.

Fully automated adaptive multi-criteria planning for cervical cancer brachytherapy, according to EMBRACE II protocol aims, resulted in improved total dose delivery (EBRT +BT) compared to clinical manual planning. Plan generation took 3-6 minutes with virtually zero workload.