Medical professionals require years of training before they can describe ultrasound images of developing foetuses. Dr Mohammad Alsharid and colleagues from the Institute of Biomedical Engineering and Nuffield Department of Women’s and Reproductive Health at the University of Oxford suggest that this task could one day be carried out by machine learning algorithms. In their latest study, the team showed how neural networks, trained by the expert knowledge of real sonographers, could convert subtle features within the images into accurate, readable captions.