Lopata Hall, Room 101
Low-dimensional Manifold Models for Image Registration and Bayesian Statistical Shape Analysis
Investigating clinical hypotheses of diseases and their potential therapeutic implications based on large medical image collections is an important research area in medical imaging. Medical images provide insights about anatomical changes caused by diseases without harmful side effects; hence is critical to disease diagnosis and treatment planning. Characterization and quantification of the anatomical changes poses computational and statistical challenges due to the high-dimensional and nonlinear nature of the data, as well as a vast number of unknown model parameters. In this talk, I will present efficient, robust, and reliable methods to address these problems. My approach entails developing a low-dimensional shape descriptor to represent anatomical changes in large-scale image data sets, and novel Bayesian machine learning methods for analyzing the intrinsic variability of high-dimensional manifold-valued data with automatic dimensionality reduction and parameter estimation. The potential practical applications of this work beyond medical imaging include machine learning, computer vision, and computer graphics.
Miaomiao Zhang is an assistant professor in the Department of Computer Science and Engineering at Lehigh University. Her research work focuses on developing novel models at the intersection of statistics, mathematics, and computer engineering in the field of medical and biological imaging. Before joining Lehigh University, Miaomiao completed her Ph.D. degree in Computer Science at University of Utah. After that, she was a postdoctoral associate at Computer Science and Artificial Intelligence Laboratory, Massachusetts Institute of Technology. She received the MICCAI Young Scientist Award 2014 and was a runner-up for the same award at MICCAI 2016.