Improving knowledge-based treatment planning for lung cancer radiotherapy with automatic multi-criteria optimized...

Area of Science:

• Medical Physics
• Radiation Oncology
• Artificial Intelligence in Medicine

Background:

• Knowledge-based planning (KBP) is an automated radiotherapy planning method using training plans to predict optimization objectives.
• KBP's effectiveness hinges on the quality of its training data.
• Integrating new automated planning systems into clinical workflows requires careful consideration of training data generation.

Purpose of the Study:

• To evaluate the impact of training data source on KBP model performance.
• To compare KBP models trained with manually created versus automatically generated multi-criteria optimized (MCO) plans.
• To assess the clinical utility and potential improvements in organ-at-risk (OAR) sparing offered by MCO-trained KBP models.

Main Methods:

• Two RapidPlan (KBP) models were created using 30 locally advanced non-small cell lung cancer patients: one with manually created clinical plans (RP_CLIN) and one with fully automatic MCO plans (RP_MCO).
• Model performance was validated on 15 patients, comparing dose-volume parameters and normal tissue complication probabilities (NTCP).
• An oncologist conducted a blind comparison of clinical (CLIN), RP_CLIN, and RP_MCO plans.

Main Results:

• RP_MCO demonstrated superior OAR sparing for the heart and esophagus compared to RP_CLIN.
• RP_MCO resulted in a 0.9% average reduction in 2-year mortality risk and a 1.6% reduction in acute esophageal toxicity risk.
• An oncologist preferred RP_MCO plans for 8 patients and CLIN plans for 7, with no preference for RP_CLIN plans.

Conclusions:

• Training KBP models with automatically generated MCO plans (RP_MCO) significantly improves OAR sparing and NTCP compared to using manually created clinical plans (RP_CLIN).
• The RP_MCO approach was selected for clinical implementation due to its superior performance.
• Optimizing training library plans during KBP model creation is crucial for enhancing the quality of future treatment plans.