2D-4D Active Appearance Models

Methodological development of 2D-4D Active Appearance Models for cardiovascular image segmentation

LUMC and University of Iowa
S.C. Mitchell, B.P.F.Lelieveldt, J.G. Bosch, R. v.d. Geest, M. Sonka, J.H.C. Reiber


Model driven image segmentation is a topic receiving widespread interest in the medical image processing community, because in the medical imaging domain, a-priori knowledge about the organ of interest and its image appearance is often indispensible to develop robust segmentation methods. In 1999, prof. Milan Sonka visited the Division of Image Processing as a Boerhaave visiting professor. During this stay, a project was started to investigate the clinical potential of Cootes's Active Appearance Models, because they represent shape and intensity knowledge to limit the shape and Models for clinical image segmentation problems. Since then, various new developments, extensions and novel applications of Active Appearance Models have been realized within this Leiden-Iowa collaboration, such as 2D hybrid AAMs (MRI), intensity normalized echocardiogram AAMs, 2D + time Active Appearance Motion Models (cardiac MRI and echo), and 3D AAM's (cardiac MRI and echo).


To develop robust image segmentation approaches for clinically challenging segmentation problems, in image modalities that are hampered by noise, inhomogeneities and acquisition artifacts.


The project is ongoing.




MRvar1   MRvar2  MRvar3  Small beating heart


For further information, please contact:
B.P.F. Lelieveldt PhD,
Division of Image Processing
Department of Radiology, 1-C2S
Leiden University Medical Center
P.O. Box 9600
2300 RC Leiden
The Netherlands
Tel. +31 (0)71 526 2285
Fax. +31 (0)71 526 6801
e-mail: B.P.F.Lelieveldt@lumc.nl

For the University of Iowa:
Milan Sonka
Dept. of Electrical and Computer Engineering
The University of Iowa
Iowa City IA 52242, USA
phone: (319)-335-6052,
fax: (319)-335-6028,