ESTRO 2025 - Abstract Book

S3067

Physics - Image acquisition and processing

ESTRO 2025

Training and validation losses converged to ~0.2 for all models in 50 epochs. All models reduced MAE and RMSE for ADC in the sub-sampled diffusion data (Fig. 1). Example ground truth, model input and output and respective difference images for SD1 and AD1 are displayed . MAE for the AD models were lower than for the SD models (Fig. 1). Results from prostate from one patient shows that there is no difference in the ADC between the models (Fig.2). Conclusion: DL models can generate high-quality images from subsampled diffusion data, enabling DWI in biologically adaptive MR-guided radiotherapy workflows. The trade-off between acceleration and image quality requires further analysis regarding repeatability and scoring by experienced readers.

Keywords: DWI, Accelerated acquisition, Deep learning

References: 1. Van Houdt PJ et. al. Eur J Cancer. 2021;153:64-71. 2. Fernando N et. al. Phy IRO. 2024; 30:100570 3. Zormpas-Petridis K et. al. Radiol Artif Intell. 2021;3(5):e200279. 4. Kawamura M et. al. Magn Reson Med Sci. 2021;20(1):99-105.

3879

Digital Poster Motion determination of lung tumours based on cine-MR images Sewa Surdashi 1 , Nils Olovsson 2,3 , Elias Lindbäck 1,4 , Eva Onjukka 1,4 , Kenneth Wikström 2,5

1 Department of Nuclear Medicine and Medical Physics, Karolinska University Hospital, Stockholm, Sweden. 2 Department of Immunology, Genetics and Pathology, Uppsala University, Uppsala, Sweden. 3 The Skandion Clinic, The Skandion Clinic, Uppsala, Sweden. 4 Department of Oncology Pathology, Karolinska Institute, Stockholm, Sweden. 5 Department of Medical Physics, Uppsala University Hospital, Uppsala, Sweden Purpose/Objective: Targeting of lung tumours in radiotherapy is challenged by tumour motion due to breathing. Traditionally, 4DCT has been employed for this purpose; however, it offers a simplistic representation of the breathing cycle, particularly in patients with irregular breathing. Cine-MR imaging facilitates an improved motion detection since it allows longer and more consistent motion monitoring without any radiation dose. This method has the potential to enhance