The development and use of machine learning systems in the healthcare sector promises significant increases in efficiency and quality. In particular, the field of radiology is very advanced. However, machine learning systems require a huge amount of datasets, the models are considered as black boxes and can detect patterns in the data that humans cannot name in advance. These characteristics seem to collide especially with the principles of data minimisation, transparency and purpose limitation enshrined in the General Data Protection Regulation (GDPR). In this presentation, Lea Köttering will analyse to what extent the development and use of machine learning systems in radiology is compatible with the GDPR.