Efficient knowledge-based planning model construction in institutions with limited cases using plan quality metrics.

Area of Science:

• Radiation Oncology
• Medical Physics
• Cancer Treatment

Background:

• Prostate cancer Volumetric Modulated Arc Therapy (VMAT) planning requires high-quality RapidPlan models.
• Limited case numbers pose challenges for constructing effective RapidPlan models.

Purpose of the Study:

• To compare the effectiveness of case selection based on Plan Quality Metrics (PQM) versus Adjusted PQM (APQM) for building RapidPlan models.
• To evaluate RapidPlan model performance using different case selection strategies.

Main Methods:

• Retrospectively scored 90 VMAT plans using PQM and APQM.
• Constructed 12 RapidPlan models using 30 cases each, with varied PQM/APQM score combinations.
• Evaluated six representative models, including P_H (top 30 PQM) and AP_H (top 30 APQM), on ten independent cases.

Main Results:

• The P_H model achieved significantly higher PQM scores and demonstrated superior consistency and organ-at-risk (OAR) sparing compared to other models.
• The AP_H model showed inconsistent performance.
• The P_H model maintained stable quality in challenging cases, outperforming manual and APQM-based models.

Conclusions:

• Case selection based on actual clinical plan quality (PQM) is more effective for building robust RapidPlan models than theoretical dose distributions (APQM), particularly with limited data.
• This PQM-based approach is practical for smaller institutions and can enhance knowledge-based planning.
• Further standardization of PQM criteria and optimization of priority settings can improve model generalizability and clinical utility.