Improving plan quality in cervical cancer brachytherapy using knowledge-based planning for direction modulated...

The bladder and rectal toxicities in cervical cancer brachytherapy are positively correlated with the DVH parameter: D2cc. This study evaluates the feasibility of knowledge-based planning to predict the D2cc, identify suboptimal plans, and improve the plan quality with Direction Modulated Brachytherapy (DMBT) applicators using knowledge-based planning based on linear relationship between overlap distances and D2cc.

The overlap volume histogram (OVH) method was used to determine the distances for 2 cm³ of overlap between the Organs at Risks (OAR) and High-Risk Clinical Target Volume (CTV_HR). Linear plots were utilized to model the OAR D2cc and 2 cm³ overlap distances. Two datasets from 45 patients (125 plans) were used to create 2 independent models: Model 1 from 59 Intracavitary (IC) and Model 2 from 66 Intracavitary-Interstitial (ICIS) plans. Performances were compared using 5-fold cross-validation. The predicted D2cc values were used as the maximum constraints in the inverse planning optimization.

The mean bladder D2cc decreased by 4.3% and 10.3% for conventional applicators, and 4.4% and 3.6% for DMBT applicators for Models 1 and 2, respectively. The rectum D2cc decreased by 3.4% and 10.7% for conventional and 3.0% and 5.0% for DMBT applicators, respectively. The sigmoid D2cc decreased by 3.1% and 6.9% for conventional and 3.2% and 5.9% for DMBT applicators, respectively. There were also significant reductions for the recto-vaginal (RV-RP) point and posterior-inferior border of symphysis (PIBS) reference points: PIBS+2cm, PIBS+1cm, PIBS-1cm, and PIBS-2cm, for both models as well.

A knowledge-based planning method successfully predicted D2cc and optimized brachytherapy plans for cervical cancer. The proposed model demonstrates the feasibility of predicting D2cc, detecting suboptimal plans, and improving the plan quality especially for DMBT where cumulative clinical experience is limited.