Shape constrained CNN for segmentation guided prediction of myocardial shape and pose parameters in cardiac MRI

•We present a CNN for direct prediction of myocardial shape and pose parameters•The parameters are linked to an underlying landmark-based statistical shape model•Semantic segmentation is used to guide robust parameter prediction and shape fitting•Consistency between predicted parameters and joint segmentation is enforced by the loss function•Regional myocardial properties can be obtained immediately from the fitted shape model