2125 - Leading the Way: Integrating Personalized 3D Cine MR Imaging into Conventional Radiotherapy Planning
Presenter(s)
T. Kim1, Y. Yoon2, S. Marasini3, S. Hollingsworth3, H. M. Gach3, M. Schmidt1, J. Park4, and P. Samson3; 1Wash U School of Medicine, Department of Radiation Oncology, St. Louis, MO, 2Department of Radiation Oncology, Yonsei Cancer Center, Heavy Ion Therapy Research Institute, Yonsei University College of Medicine, Seoul, Korea, Republic of (South), 3WashU Medicine, Department of Radiation Oncology, St. Louis, MO, 4Department of Medical Physics, Memorial Sloan Kettering Cancer Center, New York, NY
Purpose/Objective(s): Cine MRI is highly valuable in MR-guided radiology for capturing dynamic anatomical motion in real time, unlike 4D-CT and 4D-MRI. However, its widespread adoption in conventional radiotherapy faces two major barriers: 1) Radiotherapy software is currently unable to use the dynamic data; and 2) Scan time and resolution constraints limit its spatiotemporal resolution to 2D imaging, thus impairing effective 3D motion management. To overcome these challenges, we pioneered the development and integration of personalized 3D cine MR imaging into conventional radiotherapy planning.
Materials/Methods: With IRB approval, we conducted cine MRI sessions involving eight abdominothoracic cancer patients needing motion management in surface-guided radiotherapy. We leveraged advanced deep-learning techniques to create personalized 3D cine MR imaging through super-resolution reconstruction, transforming low spatial resolution 3D cine MRIs into highly detailed images. Then, we developed an innovative 3D cine viewer within the radiotherapy planning system to integrate 3D cine MRI and accurately capture comprehensive 3D respiratory motion information via auto-segmentation tools.
Results: Our project achieved three critical milestones: 1) Development of a personalized super-resolution (pSR) network using transfer learning to enhance the resolution of real-time 3D low-resolution cine MRI, yielding sharper and more detailed images. 2) Creation of an auto-segmentation (AS) network using open-source tools, capable of including 25 organs-at-risk (OARs). This network significantly speeds up and enhances the accuracy of segmentation on extensive 3D MRI data, effectively capturing dynamic anatomical motion. 3) Introduction of the 3D Cine Viewer, pioneering software integrated into a commercial Treatment Planning System (TPS). This tool features dynamic 3D MRI viewers, 3D contour viewers, and comprehensive contouring tool management, facilitating advanced manipulation and visualization of 3D cine MRI for radiotherapy planning.
Conclusion: We have successfully addressed the two core limitations of 2D cine MRI in conventional radiotherapy by incorporating personalized 3D cine MR imaging. Leveraging advanced deep-learning techniques and innovative visualization tools, we developed a personalized super-resolution network, an auto-segmentation network, and the 3D Cine Viewer. These advancements enable the dynamic capture and assessment of 3D respiratory motion, greatly enhancing radiotherapy planning and patient motion management.