Shape Modeling of the Multiobject Organ Heart

R. Pilgram, K. Fritscher (Austria), T. Fletcher (USA), and R. Schubert (Austria)


Cardiac Modeling, Cardiac Image Analysis, StatisticalShape Analysis, Medial Representation, Multi Objects


Heart diseases result in or from morphologic variations; therefore analysis of shape variability is important for diagnostic classification and understanding of biological processes. In a pilot study we analyzed 16 individual human hearts, the common shape of the main chambers and of the combined structure including their interindividual variations for a fixed time phase. The electrocardiogram triggered MRI scans were segmented semi-automatically to derive an object ensemble containing the left and right ventricle, the left and right atria, and the pericardium for each individual heart. These objects were modeled using a medial based representation method providing interindividual shape correspondence due to object intrinsic coordinate system. Based on this concept of correspondence a common shape model was generated both for the single object and for the object ensemble. The interindividual variations were analyzed using an extended PCA method showing that almost 80% of variations are lying within the first 5 modes for both, the single and the combined structures. These so far results promise the method to be a big value in quantifying interindividual shape changes of clinical relevant populations and may serve as a potential tool for segmentation, classification, and diagnosis.

Important Links:

Go Back