ESTRO 2025 - Abstract Book

S2461

Physics - Autosegmentation

ESTRO 2025

Conclusion: We trained models to automatically delineate OARs with superior contrast in MRI using CT, MRI or both. The improved performance of the MRI-based model for the hippocampi in the external validation suggests the inclusion of MRI for designated structures, even with limited training datasets. However, the overall comparable results between the different approaches suggest the potential of automatic CT-based segmentation to enable broadly applicable models. The combination CT+MRI did not result in further improvements.

Keywords: OAR, brain, imaging

References: [1] Isensee, F., Jaeger, P.F., Kohl, S.A.A. et al. nnU-Net: a self-configuring method for deep learning-based biomedical image segmentation. Nat Methods 18, 203–211 (2021). https://doi.org/10.1038/s41592-020-01008-z