国家数学与交叉科学中心合肥分中心报告

报告题目：Intrinsic Shape Analysis for Computational Neuroanatomy

报告人：Yonggang Shi, University of Southern California

时  间：2015年4月2日    下午2:30―4:00

地  点：东区管理科研楼  365英国上市官网1308室

内容提要：
In this talk, I will present our recent work on using intrinsic geometry for the automated analysis of brain imaging data. The key idea in our method is the use of Laplace-Beltrami (LB) eigenfunctions for modeling brain shapes, such as hippocampus and cortex. These tools have the advantage of being invariant to pose and scale variances, and robust to deformations from development and pathology. Using the LB eigenfunctions and topology-preserving evolution, we have developed a robust approach for surface reconstruction from segmented masks. This method can remove outliers while accurately retaining volume information. For the challenging problem of cortical surface reconstruction, we have developed a unified approach for the joint correction of geometric and topological outliers with the Reeb graph of LB eigenfunctions. By using the LB embeddings of surfaces, we have developed a novel and general approach for surface mapping via the optimization of their conformal metrics. Based on these cutting-edge algorithms for image and shape analysis, completely automated workflows have been created for the large scale analysis of brain morphometry. In our current research, these intrinsic modeling techniques are being extended to multimodal image analysis for the more accurate and robust mapping of brain structure and function. Using the reconstructed cortical surfaces, we have developed more accurate ways of normalizing cerebral blood perfusion (CBF) images with cortical thickness and area, and successfully applied them to map sex differences in brain development. For the analysis of brain connectivity, we developed a novel algorithm for fiber orientation distribution (FOD) reconstruction that can be applied to diffusion imaging data collected from a wide range of acquisition schemes. With the help of FODs and intrinsic analysis, we are able to automatically extract fiber bundles with significantly improved details and robustness using the state-of-the-art data from the Human Connectome Project.

报告人简介：
Dr. Yonggang Shi received his Bachelor and Master degree in Electrical Engineering from the Southeast University of China in 1996 and 1999, respectively. He received his Ph.D. in Electrical Engineering from Boston University in 2005. From 2005 to 2009, he was a PostDoctoral fellow at the Laboratory of Neuro Imaging (LONI) at UCLA. He was promoted to Assistant Professor at LONI in 2009. In July 2013, Dr. Shi was recruited to University of Southern California (USC) as a tenure-track Assistant Professor of Neurology and Electrical Engineering. He joins USC along with other faculty members that previously had formed the Laboratory of Neuro Imaging (LONI) at UCLA to found the newly established Institute for Neuroimaging and Informatics (INI) at USC. Dr. Shi is an Associate Editor of IEEE Transactions on Image Processing. Dr. Shi was a winner of student paper competition at the 2005 ICASSP for his work on a fast level set algorithm. He also won the Best Paper Award at the 2008 MMBIA for his work on using Reeb graphs of LB eigenfunctions to construct shape skeletons.

欢迎广大师生参加！