Professor Tim Cootes, University of Manchester
Tim Cootes recieved the BSc degree with honors in mathematics and physics from Exeter University, England, in 1986, and the PhD degree in engineering from Sheffield City Polytechnic in 1991. He began work in computer vision at the University of Manchester in 1991, where he developed an interest in statistical models of shape and appearance. He is currently a Professor of Computer Vision at the University of Manchester, with a particular focus on the construction and use of models of appearance and their application to the medical domain. His techniques for shape modelling and matching, including the Active Shape Model and the Active Appearance Model, have been adopted by many groups world-wide.
Title: Image Segmentation using Statistical Shape Models
Abstract
Many objects of interest in images can be represented as deformed versions of some average structure – for instance faces, bones and many organs in medical images. This talk will describe methods of constructing statistical models of the variation in shape and appearance of such objects from annotated sets of examples. I will give an overview of efficient regression-based techniques for matching such models to new images, including Active Appearance Models and a powerful approach which uses Random Forest regression to vote for the most likely position of each model point.
The talk will conclude with an overview of “Groupwise Registration” algorithms, which can automatically compute correspondences across large sets of 2D and 3D images. Such methods can be used to automatically construct shape and appearance models with minimal manual intervention.
———————————————————————————————————————————————————————————————————
Dorin Comaniciu, Ph.D., Siemens Corporate Technology, Princeton NJ, USA
Dorin Comaniciu is the Head of Imaging and Computer Vision at Siemens, leading one of the industry’s largest imaging research groups. He has global responsibility to oversee imaging research and transfer advanced technology and solutions to Siemens businesses. His team’s work resulted in life-saving clinical imaging products, covering scanning equipment, diagnostic imaging, and image-guided interventions. He is coauthor of the book Marginal Space Learning for Medical Image Analysis (Springer), holds 170 patents, and has coauthored 275 peer reviewed publications in the imaging field, which received 24,000 citations. Dr. Comaniciu is a Fellow of the IEEE, Fellow of the American Institute for Medical and Biological Engineering, and Top Innovator of Siemens. He is recipient of multiple awards in imaging, including the 2010 IEEE Longuet-Higgins Prize for fundamental contributions to computer vision. He graduated from University of Pennsylvania – The Wharton School, Rutgers University, and Polytechnic University of Bucharest.
Title: Shaping the Future through Innovations: Towards Image-based Personalized Medicine
Abstract
Personalized computer models of the human anatomy and physiology have the potential to improve the patient’s risk stratification and therapy selection, guidance and follow-up. The first steps to build patient-specific models involve the quantification and understanding of image data, relying on both machine learning and advanced 3D visualization. The resulting measurements are then used to solve inverse problems to estimate the physiological parameters. As an example, we will showcase personalized cardiac models that capture the patient’s heart anatomy, dynamics, hemodynamics, electrophysiology, and electromechanics. The focus is on structural heart disease and heart failure. We detail various strategies for model fitting from patient’s data and the associated statistics. We highlight example applications that make today a difference in hospitals and extrapolate on the potential of imaging technology’s potential, expectations for the future, and the increased demand for multidisciplinary projects and applications.