막수 포럼 초청 세미나
일시: 2013년 10월 11일 (금), 10:30~12:00
장소: 전산학과(E3-1) 4층 오상수 영상강의실
High-Resolution 4D-CT for Lung Cancer Image Guided Radiotherapy
Presenter: Guorong Wu (IDEA research lab, University of North Carolina, Chapel Hill)
In this talk, I will present my recent work on novel approaches for improving the accuracy of dose delivery in the image-guided lung cancer radiotherapy. Currently, 4D-CT, which is able to provide full lung/tumor motion by acquiring 3D image at each respiratory phase, has been popularly used to determine the margin for radiation fields during the treatment planning. However, due to considerable radiation dose, generally only a free-breathing 3D-CT is acquired at each treatment day for patient setup, which is done by registration with other free-breathing 3D-CT taken at the planning day. However, without knowing the exact tumor motion at the treatment day, the accuracy of patient setup may be limited. Thus, to ensure the targeting of tumor, a large treatment field margin still has to be applied. Although it is well known that a small margin of radiation fields can lead to less treatment complications, the margin cannot be safely reduced without knowing lung/tumor motion at the treatment day. I will describe our work to improve dose delivery in the image-guided lung cancer treatment by innovatively solving the challenging problem of reconstructing the 4D-CT (and further estimating 4D lung/tumor motions) from a single free-breathing 3D-CT acquired at the treatment day. In this way, we can more precisely localize the moving tumor from the reconstructed treatment day 4D-CT.
Dr. Guorong Wu received his Ph.D. degree in computer science and engineering from Shanghai Jiao Tong University, Shanghai, China. He is currently with Image Display, Enhancement, and Analysis Research Laboratory, The University of North Carolina at Chapel Hill. His research interests are image registration and machine learning in medical image analysis. He has published more than 70 papers in the top-tier journals and conferences. Also, he has developed 9 image processing software, which are widely used in neuroimaging community with >10,000 downloads from >20 countries. Especially, he organized the MICCAI workshop on Machine Learning in Medical Imaging (MLMI) in 2013.
Learning-based Hippocampus Segmentation of Hippocampus in 7.0T MR images
Presenter: Minjeong Kim (IDEA research lab, University of North Carolina, Chapel Hill)
Recent emergence of 7.0T MR scanner sheds new light on the study of hippocampus by providing much higher image contrast and resolution. However, the new characteristics shown in 7.0T images, such as richer structural information and more severe intensity inhomogeneity, raise serious issues for the extraction of distinctive and robust features for accurately segmenting hippocampus in 7.0T images. On the other hand, the conventional image features (e.g., Haar), which were designed for 1.5T and 3.0T images, generally fail to be effective, because of the considerable image artifacts in 7.0T images. In this talk, a novel learning-based method for segmenting hippocampus from 7.0T images based on improved auto-context model in a multi-atlases framework. Furthermore, the concept of unsupervised deep learning is further investigated to learn the hierarchical feature representation directly from the pre-observed image patches in 7.0T images. Promising hippocampus segmentation results were obtained on 20 7.0T images, demonstrating the enhanced discriminative power achieved by our learning-based method.
Dr. Minjeong Kim received her Ph.D. degree in computer science and engineering from Ewha Womans University, Seoul, Korea. She is currently working at IDEA research lab in University of North Carolina, Chapel Hill as a research associate. She has published a number of papers regarding image registration, segmentation as well as machine learning approaches in medical image analysis and computer aided diagnosis. She also served as a reviewer for top-tier journals and conferences as well as a committee member for international workshops in medical image analysis society.