ESTRO 2026 - Abstract Book PART II

S2108

Physics - Image acquisition and processing

ESTRO 2026

This is the first study to achieve real-time dynamic GTV prediction in lung cancer using a 3D nnU-Net model. By optimizing the key surface input region to a 100- mm radius, the model maintains high accuracy while improving computational efficiency, offering a precise, non-ionizing real-time target monitoring solution for SGRT. Keywords: Correlation model based Deep learning

Made with FlippingBook - Share PDF online