Learning-Based Compressive MRI

摘要: In the area of magnetic resonance imaging (MRI), an extensive range non-linear reconstruction algorithms have been proposed that can be used with general Fourier subsampling patterns. However, design these patterns has typically considered in isolation from rule and anatomy under consideration. this paper, we propose a learning-based framework for optimizing MRI specific anatomy, considering both noiseless noisy settings. Our learning algorithm access to representative set training signals, searches sampling pattern performs well on average signals set. We present novel parameter-free greedy mask selection method, show it effective variety rules performance metrics. Moreover also support our numerical findings by providing rigorous justification via statistical theory.