CogGen: Cognitive-Load-Inspired Fully Unsupervised Deep Generative Modeling for Compressively Sampled MRI Reconstruction
Researchers have developed CogGen, a novel framework for reconstructing magnetic resonance imaging (MRI) from compressed sensing data. This method is inspired by cognitive principles of learning, employing a staged approach to progressively incorporate data. CogGen utilizes an MRI-aware dual-threshold weighting criterion to manage k-space measurement participation, aiming to improve efficiency and reduce noise amplification compared to existing deep generative modeling techniques. AI