Related Experiment Video
Updated: Jan 9, 2026
Radiation Planning Assistant - A Streamlined, Fully Automated Radiotherapy Treatment Planning System
Published on: April 11, 2018
Development of automated whole breast 3D planning with tangential fields
Christopher Busch1, Ping Xia1,2, Chirag Shah3
1Department of Radiation Oncology, Taussig Cancer Institute, Cleveland Clinic, Cleveland, Ohio, USA.
Objectives/Purposes:
Whole breast 3D treatment planning with tangential fields often requires multiple manual adjustments, leading to inefficiency and variability. This study presents and evaluates an in-house developed automated script designed to improve planning efficiency and consistency for breast radiotherapy using tangential fields.
Materials/Methods:
19 breast cancer patients from an IRB-approved study were selected. The automated script was developed to generate beam segments through a three-step process, (1) verification of field setup, and generation of breast planning target volume (PTV), (2) duplication of the clinical plan with open tangents and (3) automatic generation of beam segments to reduce hotspot iteratively. Plans produced by the automated script (AS) were compared to the manually optimized (MO) clinical plans using eight evaluation metrics: V16Gy of the ipsilateral lung, mean heart dose, V90%, V95% (dose coverage), V105% (hotspot) of PTV, number of segments, MU (monitor unit) weighted segment area, and total MUs. A statistical analysis was performed using the Wilcoxon signed-rank test.
Results:
Four of the eight metrics showed statistical differences. For MO and AS plans, the average ± SD of ipsilateral lung 16 Gy was 15.05% ± 3.27% vs. 15.10% ± 3.26% (p = 0.04) while the mean heart dose was 0.87 Gy ± 0.52 Gy, vs. 0.87 Gy ± 0.53 Gy (p = 0.68). With regards to homogeneity and coverage, the PTV V105% were 2.0% ± 5.0% and 3.7% ± 6.4% (p = 0.03) while the PTV V90% were 98.8% ± 1.5% and 99.1% ± 0.9% (p = 0.14), V95% were 92.0% ± 4.4% and 92.4% ± 3.2% (p = 0.63) respectively. The number of segments and MU weighted segment areas were 7.3 ± 1.3 and 10.9 ± 1.9 (p < 0.001); 268.2 ± 96.6 cm2 and 262.6 ± 92.9 cm2 (p < 0.001). The total MUs was 351.7 ± 99.0 and 354.7 ± 101.2 (p = 0.05).
Conclusion:
The automated script successfully generated high-quality tangential breast plans that met all clinical constraints. It significantly improved consistency while maintaining clinical acceptability, demonstrating potential for broader use in automated 3D planning.
