Embedded Files
Jong Chul Ye is a Professor of the Kim Jaechul Graduate School of Artificial Intelligence (AI) of Korea Advanced Institute of Science and Technology (KAIST), Korea. He received the B.Sc. and M.Sc. degrees from Seoul National University, Korea, and the Ph.D. from Purdue University, West Lafayette. Before joining KAIST, he worked at Philips Research and GE Global Research in New York. He has served as an associate editor of IEEE Trans. on Image Processing, and an editorial board member for Magnetic Resonance in Medicine. He is currently an associate editor for IEEE Trans. on Medical Imaging, and a Senior Editor of IEEE Signal Processing Magazine. He is an IEEE Fellow, was the Chair of IEEE SPS Computational Imaging TC, and IEEE EMBS Distinguished Lecture . He ws a General Cochair (with Mathews Jacob) for IEEE Symp. On Biomedical Imaging (ISBI) 2020. His research interest is in machine learning for biomedical imaging and computer vision.
PROF. JONG CHUL YE
PROF. JONG CHUL YE
Korea Advanced Institute of Science and Technology (KAIST)
Score-based Diffusion Models for Inverse Problems in Medical Imaging
Score-based Diffusion Models for Inverse Problems in Medical Imaging
Abstract:
Recently, diffusion models have been used to solve various inverse problems for medical imaging applications in an unsupervised manner. However, the current solvers, which recursively apply a reverse diffusion step followed by a measurement consistency step, often produce sub-optimal results. By studying the generative sampling path, we show that current solvers throw the sample path off the data manifold, and hence the error accumulates. Furthermore, diffusion models are inherently slow to sample from, needing few thousand steps of iteration to generate images from pure Gaussian noise. In this talk, we show that starting from Gaussian noise is unnecessary. Instead, starting from a single forward diffusion with better initialization significantly reduces the number of sampling steps in the reverse conditional diffusion. This phenomenon is formally explained by the contraction theory of the stochastic difference equations like our conditional diffusion strategy - the alternating applications of reverse diffusion followed by a non-expansive data consistency step. Furthermore, we propose an additional correction term inspired by the manifold constraint, which can be used synergistically with the previous solvers to make the iterations close to the manifold. Experimental results with image inpainting, and compressed sensing MRI and sparse-view CT demonstrate that our method can achieve state-of-the-art reconstruction performance at significantly reduced sampling steps.
Shanshan Wang is a professor at the Paul C Lauterbur Research Center, Chinese Academy of Sciences, where she targets to develop novel adaptive learning methods and applications for fast medical imaging and intelligent medical analysis. With dual Ph.D. degrees in Biomedical Engineering (BME) and Computer Science (CS), she pioneered the integration of core CS methodologies with imaging sciences. Her innovative approaches led to several key contributions to the field of medical imaging: performed seminal work in introducing deep learning to MR imaging, to open up a new era of learning reconstruction for MRI and enabled “real-time” observation of numerous challenging biological structures and processes; designed unrolled iterative feature refinement framework, which could achieve sub-millimeter high-resolution MR imaging. This work was selected as “Research highlight” by the prestigious journal “Physics in Medicine and Biology”; developed a series of stat-of-the-art algorithms that can discover disease characteristics that were previously not detectable by the naked eye and release clinicians from repetitive and tedious work leading to full automation; and designed an open framework AIDE to handle imperfect training datasets having limited annotations, lacking target domain annotations, or containing noisy annotations.
Dr. Shanshan Wang has been a Gordon Plenary Lecturer, ISMRMR NIBIB New Horizons Plenary Lecturer, IEEE senior memeber, OCSMRM BoT/Life member, Deputy editor of Magnetic resoance in medicine, Associate editor of Pattern Recognition and Associate editor of Biomedical Signal Processing and Control, etc.
PROF. Shanshan wang
PROF. Shanshan wang
Paul C Lauterbur Research Center, Chinese Academy of Sciences
Deep Learning for mr reconstruction: a trend to learn more with less data
Deep Learning for mr reconstruction: a trend to learn more with less data
Abstract:
Deep learning has become a dominant approach in accelerating MR imaging. It employs deep neural networks to draw knowledges from available datasets and then use the trained networks to assist accurate image reconstruction from limited measurements. This talk briefly gives the traits and trends of these techniques, which have experienced a paradigm shift from “big data, small physics” to “small data, lots of physics”. Specially, these methods have moved from fully-supervised learning, to semi-supervised learning, to unsupervised learning. Some discussions and outlooks have also been provided.
Page updated
Google Sites
Report abuse