Atlas-Based whole-body PET-CT segmentation using a passive contour distance

Authors:
Fabian Gigengack;Lars Ruthotto;Xiaoyi Jiang;Jan Modersitzki;Martin Burger;Sven Hermann;Klaus P. Schäfers
Affiliations:
European Institute for Molecular Imaging (EIMI), University of Münster, Germany, Department of Mathematics and Computer Science, University of Münster, Germany;Institute of Mathematics and Image Computing, University of Lübeck, Germany;Department of Mathematics and Computer Science, University of Münster, Germany;Institute of Mathematics and Image Computing, University of Lübeck, Germany;Institute for Computational and Applied Mathematics, University of Münster, Germany;European Institute for Molecular Imaging (EIMI), University of Münster, Germany;European Institute for Molecular Imaging (EIMI), University of Münster, Germany
Venue:
MCV'12 Proceedings of the Second international conference on Medical Computer Vision: recognition techniques and applications in medical imaging
Year:
2012

Citing 4
Cited 0

A Multiphase Level Set Framework for Image Segmentation Using the Mumford and Shah Model

International Journal of Computer Vision
Fair: Flexible Algorithms for Image Registration

Fair: Flexible Algorithms for Image Registration
Automated registration of whole-body follow-up MicroCT data of mice

MICCAI'11 Proceedings of the 14th international conference on Medical image computing and computer-assisted intervention - Volume Part II
Active contours without edges

IEEE Transactions on Image Processing

Quantified Score

Hi-index	0.00

Visualization

Abstract

In positron emission tomography (PET) imaging, the segmentation of organs is necessary for many quantitative image analysis tasks, e.g., estimation of individual organ concentration or partial volume correction. To this end we present a fully automated approach for wholebody segmentation which enables large-scale and reproducible studies. The approach is based on joint segmentation and atlas registration. The classical active contour approach by Chan and Vese is modified to a novel passive contour energy term with implicitly incorporated information about shape and location of the organs. This new energy is added to a registration functional which is based on both functional (PET) and morphological (CT) data. The proposed method is applied to medical data, given by 13 PET-CT data sets of mice, and quantitatively compared to manually drawn VOIs. An average Dice coefficient of 0.73 ± 0.10 for the left ventricle, 0.88 ± 0.05 for the bladder, and 0.76 ± 0.07 for the kidneys shows the high accuracy of our method.