ESTRO 2024 - Abstract Book

S4507

Physics - Machine learning models and clinical applications

ESTRO 2024

Figure 2. Box plots for dose parameters derived from deep learning-based dose prediction (AI Prediction) and the automated plan optimization (Auto Plan). The mean dose to the structure is presented on the left, and the volume of different dose levels on the right.

Conclusion:

We developed and evaluated a clinical implementation of deep learning-based dose prediction and automated planning. The clinical implementation was feasible and generalized well to data from another RT center. Further development is needed to extend the framework to other treatment planning systems and to improve the performance of automatic optimization. The clinical implementation of this method has the potential to facilitate and improve RT dose planning for breast cancer patients and shows potential for multi-center collaboration, treatment harmonization, and plan optimization.

Keywords: Dose prediction, auto-planning, breast cancer