Readings in Vision and Graphics
edited by Luc Van Gool, Gábor Székely, Markus Gross, Bernt Schiele
Predictive Properties of Statistical Shape Models
2011. XVI, 102 pages. EUR 64,00.
Shape modeling and prediction from incomplete observation is a key problem in numerous clinical applications. The specific aim of this work is to develop a generic prediction framework which takes into account the main sources of uncertainty within the prediction chain and provides a quantitative evaluation of the prediction accuracy.
The quality of statistical shape model utilized to learn variability of the target organ shape is influenced by the quality of the underlying dense correspondence between the individual shapes of the training set. We propose an extension to correspondence establishment over a population based on the minimization of the description length function, which allows considering objects with arbitrary topology.
Furthermore, the representativeness of the training data as well as the quality of the available observation is important for reliable predictions. We propose a framework which is able to take both related uncertainties into account and allows the derivation of a distribution of probable shapes given a noisy observation. The reliability of the prediction is quantified by the confidence region around the predicted shape and can be conveniently visualized.
About the author
Ekaterina Mishina obtained her M.Sc. (Dipl.) degree
in Mathematics and Applied Mathematics from the Lomonosov Moscow State
University, Russia, in 2003. After graduation she was working for two years as
a software engineer in a company developing microprocessors. From 2006 to 2010,
she was research assistant and Ph.D. student at the Computer Vision Laboratory
Keywords: shape modeling, statistical shape models, prediction accuracy, topology, organ´s shape.
Direkt bestellen bei / to order directly from: Hartung.Gorre@t-online.de
Hartung-Gorre Verlag / D-78465 Konstanz / Germany
Telefon: +49 (0) 7533 97227 Telefax: +49 (0) 7533 97228