Accelerated and quantitative three-dimensional molecular MRI using a generative adversarial network.

GAN-ST can substantially reduce the acquisition time for quantitative semi-solid MT/CEST mapping, while retaining performance even when facing pathologies and scanner models that were not available during training.