Comparing tumor burden-based classification with traditional methods for knowledge-based VMAT planning in multiple...

Area of Science:

• Radiation Oncology
• Medical Physics
• Radiotherapy Planning

Background:

• Accurate dose prediction is crucial for effective volumetric modulated arc therapy (VMAT) in treating multiple brain metastases (multi-BM).
• Knowledge-based planning (KBP) offers potential for improving treatment plan quality and consistency.

Purpose of the Study:

• To systematically compare the dose prediction performance of three distinct knowledge-based RapidPlan models: Broad-Spectrum Model (Model-BS), K-means Clustering-Based Classification Model (Model-CA), and Volume Stratification Model (Model-V).
• To evaluate these models in the context of VMAT planning for patients with multi-BM.

Main Methods:

• Retrospective analysis of 138 multi-BM patients treated with VMAT (30 Gy/5 Fr).
• Development of Model-BS (general), Model-CA (clustering based on lesion count/volume), and Model-V (volume stratification).
• Comparison of automated plans (BS-plans, CA-plans, V-plans) against manual plans (MP-plans) using dose metrics (CI, GI, OAR doses) in a test group of 16 patients.

Main Results:

• All KBP models demonstrated comparable or improved target coverage versus manual plans.
• Model-CA and Model-V significantly reduced high-dose volumes in the brain compared to Model-BS (p < 0.05).
• Model-V achieved the lowest gradient index (GI=5.55) and highest conformity index (CI=0.924), while Model-CA showed a better balance for OAR sparing (e.g., optic nerve Dmax).

Conclusions:

• Classification-based KBP models, utilizing tumor lesion count and total volume, significantly enhance dose distribution for multi-BM VMAT.
• Model-V excels in dose gradient optimization, whereas Model-CA provides a superior overall balance of target conformity, dose fall-off, and organ-at-risk sparing.