Related Experiment Video
Updated: Apr 14, 2026
Radiation Planning Assistant - A Streamlined, Fully Automated Radiotherapy Treatment Planning System
Published on: April 11, 2018
Development and feasibility assessment of an automated, synthetic CT-guided plan selection system for bladder cancer
Huipeng Meng1, Xinrui Wang1, Pengfei Liu1
1Radiotherapy Center, Tianjin First Central Hospital, School of Medicine, Nankai University, Tianjin, China.
Background:
Interfractional bladder volume variation during radiotherapy challenges the optimality of a single treatment plan. We developed and conducted a pilot feasibility study of a decision support system to select the optimal plan of the day (PotD) using daily imaging, aiming to translate a sophisticated online adaptive radiotherapy (ART) workflow into clinical practice.
Methods:
A Python script was developed in the RayStation treatment planning system (TPS). For12 patients, a library of three volumetric modulated arc therapy (VMAT) plans was auto-generated from planning computed tomography (CTs) representing small, medium, and large bladders. For daily adaptation, a synthetic CT (sCT) was generated from the pre-treatment cone beam computed tomography (CBCT) using a corrected deformable image registration (DIR) methodology [fixed CBCT, moving planning CT (pCT)] with a hybrid deformable registration algorithm. All library plans were recalculated on the sCT. A transparent linear scoring function, based on 15 dose-volume histogram (DVH) metrics, was used to identify the optimal plan. This workflow was retrospectively applied to 24 patient fractions. To validate the system's geometric accuracy, generated sCTs were compared against physician-verified ground truth contours using Dice similarity coefficient (DSC) and Hausdorff distance (HD). Clinical utility was assessed by comparing the automated plan selection against a blinded senior radiation oncologist's decision.
Results:
The sCTs demonstrated high Hounsfield unit (HU) fidelity. Geometric validation confirmed high accuracy, with a mean PTV DSC of 0.908±0.010 and a 95% Hausdorff distance (HD95) of 2.67±0.54 mm. In the 24 analyzed fractions, the medium bladder plan (P2) was selected in 95.8% of cases and achieved a significantly higher mean score (85.8±4.9) than the other plans (P<0.001). In the clinical utility assessment, the automated system achieved a high concordance rate of 87.5% (21/24 fractions) with the human expert. Discordant cases reflected the system's objective adherence to pre-defined scoring criteria, balancing optimal organ at risk (OAR) sparing with rigorous target coverage protocols.
Conclusions:
The developed CBCT-based decision system demonstrates technical feasibility and high geometric fidelity in automatically selecting an optimal daily plan by adapting to interfractional anatomical changes. This approach represents a feasible and practical method for implementing online ART for bladder cancer patients in a routine clinical setting.
