ESTRO 2024 - Abstract Book

S3499

Physics - Dose prediction, optimisation and applications of photon and electron planning

ESTRO 2024

difference in quality compared to fully Pareto-optimized plans. The combination of KNN and DA yielded the best results. The analysis highlighted the strong association between rectum dose and rectum volume overlapping the PTV, offering insights into potential further optimization.

Conclusion:

The systematic auto-generation of Pareto-optimal plans in hypo-fractionated prostate radiotherapy reveals the potential to reduce organ-at-risk doses, particularly for the rectum, while maintaining target coverage. Machine learning algorithms, trained with Pareto data, can replicate high-quality plans effectively. This study highlights the importance of systematically generated data in training machine learning models to achieve plan quality beyond current clinical standards.

Keywords: Pareto optmization, machine learning,

References: