Stage Master 2: Parameter study for cardiac segmentation

Context. Cardiac segmentation with elastic deformable templates has been studied for years, in particular at Creatis. Although still evolving, the method and its implementation are now able to provide good results in 3D [2] and 2D+t [3]. However, fine-tuning the segmentation parameters to fit patient data is still cumbersome due to the variety of acquisition sequences, image resolution and cardiac pathologies. Although the segmentation method under study has only a reduced number of parameters, their effect is difficult to predict and in practice, obtaining good myocardium segmentations is still an art mastered by a few. In the context of the Gwendia ANR project, we are aiming at using grid computing to enable massive processing of cardiac images. In particular, workflows have been developed to execute parameter sweep experiments for cardiac segmentation on the EGEE grid [1] with the goal to help parameter tuning.