ESTRO 2024 - Abstract Book

S3762

Physics - Image acquisition and processing

ESTRO 2024

2. Blur perception assessment

For the BH dataset, our method achieves a competitive performance in terms of CPBD with 0.309 ± 0.023, which improved from 0.069 ± 0.013 (average improvement 0.24); however, the improved value was still lower than that of scanned high-quality images (0. 486± 0.031) by 0.177. For the FB dataset, our method achieves a competitive performance in terms of CPBD with 0.317 ± 0.032, which improved from 0.071 ± 0.02 (average improvement 0.246) but was still lower than that of scanned high-quality images (0.494 ± 0.035) by 0.177. Furthermore, our method not only achieves higher clarity metrics but also has a stable performance with dense distribution.

Conclusion:

In this study, we developed a personalized model for constructing personalized high-quality enhanced 4D-MRI for liver cancer radiotherapy. Quantitative evaluations revealed that the proposed method can enhance the contrast and reduce noise to the level of real high-quality MRI images. Compared to the GM, the personalized one showed statistically significant improvements. Furthermore, the blur perception assessment on edge points demonstrated the ability of our model to improve image quality for human blur perception. Thus, this method shows potential for application in clinical settings.

Keywords: 4D-MRI, personalized deep learning,

459

Proffered Paper

Generating synthetic computed tomography for radiotherapy: SynthRAD2023 challenge report

Evi Huiben 1 , Maarten Terpstra 2,3 , Arthur Jr Galapon 4 , Suraj Pai 5 , Adrian Thummerer 4 , Peter Koopmans 6 , Manya Afonso 7 , Maureen van Eijnatten 1 , Oliver Gurney-Champion 8,9 , Zoltan Perko 10 , Matteo Maspero 2,3 1 Eindhoven University of Technology, Biomedical Engineering, Eindhoven, Netherlands. 2 University Medical Center Utrecht, Radiotherapy, Utrecht, Netherlands. 3 University Medical Center Utrecht, Computational Imaging Group for MR Diagnostics & Therapy, Utrecht, Netherlands. 4 University Medical Center Groningen, Radiation Oncology, Groningen, Netherlands. 5 Maastricht University Medical Centre, Radiation Oncology, Maastricht, Netherlands. 6 Radboud University Medical Center, Radiation Oncology, Nijmegen, Netherlands. 7 Wageningen University & Research, Wageningen Plant Research, Wageningen, Netherlands. 8 Radiology and Nuclear Medicine, Amsterdam UMC, Amsterdam, Netherlands. 9 Cancer Center Amsterdam, Imaging & Biomarkers, Amsterdam, Netherlands. 10 Delft University of Technology, Radiation Science and Technology, Delft, Netherlands

Purpose/Objective:

RT is pivotal in treating cancer patients, aiming to deliver a precise radiation dose to the tumor while sparing healthy tissues. However, obtaining accurate patient anatomy for RT planning can be challenging, especially in daily adaptive radiotherapy. Additionally, even without considering daily adaptation, adopting MRI-only radiotherapy by avoiding