Synthesized 7T MRI from 3T MRI using generative adversarial network: validation in clinical brain imaging

Ultra-high field 7T MRI provides exceptional tissue contrast and anatomical details but is often cost-prohibitive and not widely accessible in clinics. A generative adversarial network (SynGAN) was developed to generate synthetic 7T images from the widely used 3T images. The synthetic 7T images achieved improved tissue contrast and anatomical details compared to the 3T images. Meanwhile, the synthetic 7T images showed comparable diagnostic performance to the authentic 7T images for visualizing a wide range of pathology, including cerebral infarction, demyelination, and brain tumor.