University of Texas at Austin

Past Event: Oden Institute Seminar

Analysis of Sheet-Like Anatomical Shapes Using Medial Representations

Paul Yushkevitch , Associate Professor, Penn Image Computing & Science Laboratory, University of Pennsylvania

3:30 – 5PM
Thursday Feb 1, 2018

POB 6.304

Abstract

Many structures in the human body are thin and flat, having what might be called a sheet-like shape. Interesting aspects of sheet-like shapes can be captured by the medial axis, a geometrical construct generated by thinning a shape until nothing but an infinitely thin skeleton remains. The medial axis captures the overall three-dimensional shape of sheet-like objects, while also giving rise to a well-formed definition of local thickness. I will discuss a set of computational modeling techniques that allow features derived from the medial axis to be used for statistical analysis of shape, such as for deriving the mean shape from a sample of anatomical objects, or characterizing the effects of disease on sheet-like organs. The key challenge addressed by these techniques is how to describe multiple exemplars of a shape in a consistent manner, such that point-wise correspondences between different exemplars’ medial axes can be found. I will present several applications of medial modeling in medical image analysis, including approaches for the segmentation, geometrical modeling, and statistical analysis of heart valves. Time permitting, I will present very recent work that uses related shape analysis techniques to solve complex groupwise image registration problems that arise in high-resolution ex vivo imaging. Bio Paul Yushkevich received his Ph.D. in Computer Science in 2003 from the University of North Carolina at Chapel Hill. After a postdoc in the Department of Radiology at the University of Pennsylvania, he joined the faculty there, and is currently Associate Professor. His research interests include statistical shape analysis, computational object representation, automatic image segmentation, groupwise image registration, and structure-specific analysis of anatomic, functional and diffusion-weighted MRI data; as well as application of these techniques to cross-sectional and longitudinal studies of brain and cardiac disorders. An area of particular interest is high-resolution ex vivo imaging and detailed characterization of the human medial temporal lobe, the seat of memory in the human brain and the site of earliest Alzheimer's disease pathology. He is active in the development of open-source image analysis software, such as ITK-SNAP, a 3D image segmentation tool that is used widely in the biomedical imaging community.

Event information

Date
3:30 – 5PM
Thursday Feb 1, 2018
Location POB 6.304
Hosted by Michael S. Sacks