Implementation and Dosimetric Evaluation of a Knowledge-Based IMRT Model for Left-Sided Breast Radiotherapy with DIBH

This study aimed to dosimetrically evaluate the plan quality of a knowledge-based planning (KBP) model developed from intensity-modulated radiation therapy (IMRT) plans for left-sided breast cancer treated with deep inspiration breath hold (DIBH), and to assess its clinical applicability.

Eighty patients previously treated with DIBH were included. A RapidPlan-based KBP model was developed using clinically approved plans (ClinP) within a Halcyon-equivalent workflow on the Ethos platform. Dosimetric performance of model-only plans (KBP) and plans refined with limited manual optimization (KBP+MP) was compared against ClinP for planning target volume (PTV) coverage and organ-at-risk (OAR) sparing.

The KBP model demonstrated satisfactory training performance, with acceptable coefficients of determination (R²) and mean squared error (MSE) values for most OARs. Both KBP and KBP+MP achieved PTV coverage comparable to ClinP, particularly for D95 and D98. In the left lung, V20 and D2% were reduced, and both the mean dose and D2% to the left anterior descending (LAD) artery were significantly lower in KBP plans, although lung V5 showed a slight increase.

Our findings support the clinical applicability of KBP. While KBP effectively reduces high-dose regions, it may increase low-dose exposure particularly in distant OARs such as the contralateral lung and breast. The KBP+MP approach provided more balanced OAR sparing while simplifying the overall planning workflow compared with manual methods.

The KBP model demonstrated clinical feasibility for left-sided breast IMRT under DIBH conditions. Model-generated plans achieved target coverage comparable to clinically approved plans while providing acceptable OAR sparing. Although KBP alone tended to increase low-dose exposure in contralateral structures, limited manual refinement improved overall dose balance. These findings support the integration of structured KBP workflows into standardized DIBH breast IMRT planning.