MICCAI 2020 Daily - Tuesday

2 Paper Presentation 10 Sina Amirrajab is a second-year PhD student at Eindhoven University of Technology (TU/e) in the Netherlands, where he is a member of Josien Pluim’s Medical Image Analysis Group. He is also part of a Marie Curie-funded European project called openGTN, which has many industrial and academic partners. His main supervisor is Marcel Breeuwer, a principal scientist at Philips and a professor at TU/e. His industrial supervisors are Cristian Lorenz and Jürgen Weese, from Philips Research Hamburg. Originally from Iran, this is Sina’s first time at MICCAI. He speaks to us ahead of his oral session today. Sina’s work is about generating images with ground truth labels that can be used for other downstream supervised tasks. His paper contains a number of different elements, including an anatomical model based on the XCAT computerized phantom; a conditional GAN architecture based on a state- of-the-art image-to-image translation network; a network which evaluates these generated or synthesized images; and a simulation part to simulate XCAT-GAN for Synthesizing 3D Consistent Labeled CardiacMR Images on Anatomically Variable XCAT Phantoms Sina Amirrajab cardiac MRI images to be able to train an initial network with a task of segmenting real images. “I always distinguish between simulating and synthesizing,” Sina explains. “Simulating is based on the physics of imaging modalities combined with a physiological or anatomical model, with access to parameters that you can vary. Synthesizing is based on using different kinds of techniques, like GANs and other generative models, to synthesize images based on what DAILY Tu e s d a y