Nonstationary image artifacts frequently arise in MRI from off-resonance and motion. Current methods to correct these nonstationary effects are computationally expensive. Stanford researchers have developed a new deep learning framework to improve image quality in minutes.