Knowledge-Based Planning for Chest Wall with Lymph Nodes Irradiation VMAT 📝

Author: Nesrin Dogan, Panagiota Galanakou, Robert Kaderka 👨‍🔬

Affiliation: University of Miami, Sylvester Comprehensive Cancer Center, University of Miami Sylvester Comprehensive Cancer Center 🌍

Abstract:

Purpose:
To develop knowledge-based treatment planning (KBP) for volumetric modulated arc therapy (VMAT) in chest wall treatments with regional nodal involvement. Given the challenges posed due to variability in target volumes and organ-at-risk (OAR) anatomy, the models aim to streamline planning and enable clinical implementation.
Methods:
Two separate KBP models were configured: 47 left-sided and 44 right-sided chest wall patients with regional lymph node involvement. Optimization objectives were refined iteratively until the plans met clinical quality, integrating model-generated, manual, and normal tissue objectives. Model performance was evaluated using R², X², and Mean Squared Error (MSE). KBP plans were validated using 10 plans per model comparing dosimetry to clinical plans using the two-tailed paired t-test (p<0.05). The KBP plans were generated without planner intervention, but further optimization was performed by adding a Monitor Unit objective.
Results:
The left-sided model significantly (p<0.05) reduced the esophagus maximum dose (-497±147cGy), ipsilateral lung dose (V1800cGy:-3.2±2.3%), and contralateral breast dose (V500cGy:-1.9±1.8%). Reductions in mean doses to thyroid (-296±96cGy) and heart (-1.8±14cGy) were observed, but differences were not statistically significant. Minimal reductions with p <0.05 in internal mammary node (IMN) coverage and an increase in contralateral lung dose (for V480cGy, p >0.05) were observed but both remained within acceptable limits. The right-sided RP model reduced the mean heart dose (-46±9cGy), thyroid (-220±85cGy), ipsilateral lung (-4.4±0.04cGy), and contralateral breast (V500cGy:-4.0±2.0%). A slight increase in the esophagus maximum dose and a reduction in IMN coverage were observed. The low OAR MSE values (<0.05) indicate reliable dosimetric predictions, even with variability in R² (0.2-0.9), which suggests varying predictive strength without overfitting. Meanwhile, X² values close to 1 confirm good model fit.
Conclusion:
Both KBP models produced high-quality plans with equivalent target coverage, though there were some tradeoffs in OAR sparing compared to the clinical plans.

Back to List