Shape-based Segmentation
Template divided into patches.
© dkfz.de
An extension to the segmentation with Active Contour Models are Active Shape Models [1]. First a template model is generated from a training data set. The template model is the surface of a segmentation. It is then divided into a certain number of patches. The division is performed according to the curvature of the contour. Points with similar curvature are grouped together. During the segmetation process the patches are adapted to the image data independently. After all patches are adapted, the final contour is generated by interpolation. The resulting algorithm is a template based approach for the segmention of organs at risk. This way the robustness of the underlying ACM approach is improved and the difficult training of a Point Distribution Model PDM is avoided.
Intermediate segmentation step (left). Final segmentation result (right).
© dkfz.de
References
[1] Cootes TF, Taylor CJ: Statistical models of appearance for medical image analysis and computer vision. Proc SPIE Med Img: 236–248, 2001.