Evaluating deformable models for segmentation of cardiac images

For diagnosis and treatment planning of heart diseases temporal series of tomographic images (EB-CT, MRI) are acquired. Their evaluation can be simplified by segmenting the images and enable the measurement of several medical indicators (ejection fraction, end-diastolic volume, end-systolic volume). This paper presents three different active contour model based techniques applied to segmentation of cardiac images. In order to find the ventricle contours different image features are used. In addition this paper discusses methods for evaluation of the accuracy of segmentation results.