Knowledge Driven Segmentation

Knowledge Driven Automatic Segmentation in Temporal and Spatial Cardiovascular Image Sequences

Hans van Assen, M.Sc


In the past decade, model guided image segmentation has become focus of medical image processing research groups, yielding to the development of snakes, point distribution models, active shape models and eventually Active Shape and Appearance Models. The latter two describe the shape and shape variations over a population as a mean shape and a number of eigenvariations, which can be extracted e.g. by principal component analysis (PCA). The extension from 2D to 3D shape modeling and matching however, is far from straightforward. In previous work, we have presented an extension of 2D Active Appearance Models to 3D, which has successfully been applied to segmentation of cardiac MR images. The purpose of this project is to develop similar modeling and matching methods for Active Shape Models in such a manner that:

  • it can be applied to sparsely sampled data of arbitrary planar orientation.
  • a limited amount of new training data is required to apply the 3D ASM's for to a variety of 3D segmentation applications, i.e. to develop a cross-modality 3D ASM.


The main goal of this project is the development of 3D ASM modeling and matching methods for segmentation of cardiac CT and MR image sequences, irrespective of the orientation of the image data.

Knowledge Driven Automatic Segmentation 1

Knowledge Driven Automatic Segmentation 2

Click on the images, to activate the movies.




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