Recent remarkable advances in learning-based methods are revolutionizing the field of image analysis resulting in a paradigm shift towards data-driven approaches. These methods have already shown tremendous success in a wide range of applications from computer vision such as object detection and object category classifications. To perform efficiently though, such methods typically require a large number of training sample and this requirement is highly impractical or impossible in applications from medical imaging where it is very laborious and expensive to generate labeled data. Another drawback of learning-based method is that they do not come with provable performance guarantees. One remedy to such limitations is to combine data-driven methods with prior knowledge through mathematical and/or physical models in such a way to optimize processing, reconstruction and analysis of medical imaging data without the need of extensive training. This workshop will bring together researchers with a different background ranging from optimization, inverse problem, numerical and harmonic analysis and machine learning to advance state-of-the-art methods combining data- and model-driven approaches for medical imaging.