Medical image computing aims at developing computational strategies for robust, automated, quantitative analysis of relevant information from medical imaging data in order to support diagnosis, therapy planning and follow-up, and biomedical research. Medical image analysis is complicated by the complexity of the data itself—involving 3D tomographic images acquired with different modalities that are based on different physical principles, each with their own intrinsic characteristics and limitations, and by the complexity of the scene—involving normal and pathological anatomy and function, with complex 3D shapes and significant inter-subject variability. Hence, model-based approaches are needed that take prior knowledge about the image appearance of the relevant objects in the scene into account. These models are parameterized to deal with variability in object appearance, such that the image analysis problem can be formulated as an optimization problem of finding the model parameters that best explain the image data. Depending on the representation chosen for the model, different approaches can be discriminated. Machine learning offers the possibility to learn suitable models from previously analyzed data itself.