Brain Magnetic Resonance Imaging Generation using Generative Adversarial Networks

Magnetic Resonance Imaging (MRI) is nowadays one of the most common medical imaging technology, due to its non-invasive nature and the many kind of supported sequences (modalities), that provide unique insights about a particular disease. However, it is not always possible to acquire all the sequences required, for several reasons such as prohibitive scan times or allergies to contrast agents. To overcome this problem and thanks to the recent improvements in Deep Learning, in the last few years researchers have been studying the application of Generative Adversarial Networks, a promising paradigm in deep learning, to generate the missing modalities. In this work we developed and trained two models of Generative Adversarial Networks, called MI-pix2pix and MI-GAN, to solve the problem of generating missing modalities for brain MRIs. In particular, our approaches are multi-input generative models, as they exploit as input several MRI modalities to generate the missing one. Our results are promising and show that the developed models are able to generate rather realistic and good quality images.