A realistic anthropomorphic numerical model of the beating heart Rana Haddad 1 , Patrick Clarysse, Maciej Orkisz, Pierre Croisille, Didier Revel, and Isabelle E. Magnin Creatis, CNRS Unit #5515, INSERM U630, F-69621 Lyon Cedex, France, rana.haddad@creatis.insa-lyon.fr, WWW home page: http://www.creatis.insa-lyon.fr/ Abstract. A realistic anthropomorphic numerical model of the beating heart is presented. It includes the main cardiac anatomical structures, vessels junctions and part of the coronary network. Its main feature is that it is based on an imaging study on the same human subject from which both structural and motion information are retrieved. This confers to the model a remarkable consistency. Heart’s deformation is assessed through successive non linear registrations in cine MR sequences. The resulting model can be used for the evaluation of cardiac image pro- cessing algorithms such as myocardium segmentation and cardiac image registration. 1 Introduction Up-to-date medical imaging modalities, such as Magnetic Resonance Imaging (MRI) are now able to provide very accurate pictures of the heart’s anatomy in 3D and through the cardiac cycle. Image acquisitions generally relies on trig- gering systems (i. e. ECG gating) to reconstruct the anatomy from data record- ing over a certain time interval (currently several cardiac cycles). For diagnostic purposes, clinicians need image post-processing techniques in order obtain quan- titative parameters, currently cavity volumes, ejection fraction or local motion indices such as myocardial wall thickening. Therefore, a significant number of methods have been developed for the segmentation of the heart and the estima- tion of its motion [1], [2], [3], [4]. Some of them may surely obtain good results in specific cases but the novel user usually lacks objective performance comparison. A beating heart model could serve as a reference to evaluate the performances of segmentation methods [5], [6]. Such a model could also be of great interest for the evaluation of reconstruction [7] and image registration [8] algorithms. In these cases, the model would include not only a computer description of the main cardiac structures but also the associated medical images in mono or even multi-modalities. Another possibility of such a model would be the training for minimally invasive beating heart surgeries [9]. If the model is doted of some tuning parameters, various normal and pathological conditions could then be simulated.