MASC: Metal-Aware Sampling and Correction via Reinforcement Learning for Accelerated MRI
MASC is a reinforcement learning approach that jointly optimizes how k-space data is sampled and how artifacts are corrected for accelerated MRI. It uses an artifact-aware policy to select which lines to acquire, while a U-Net network simultaneously learns to correct metal artifacts. This co-adaptive training significantly improves image quality compared to conventional methods.