The learning process of deep learning systems is a static procedure. The systems learn from a snapshot of the data initially available and the learned concepts and parameters remain unchanged during the test phase. Deploying these systems into dynamic environments, such as clinical settings, is prone to failure because the behavior of deep learning systems is unpredictable when facing new cases or examples not included during the training phase and because the systems cannot adapt to new clinical requirements or feedback from experts. To goal of this project is to develop dynamic learning strategies for deep learning systems in clinical environments. We focus on the development of algorithms for analysis of cardiac MR images. The project is funded by Dutch Technology foundation and is part of the DLMedIA program with Pie Medical Imaging as our industrial partner.