DA-GAN Fast-MRI

Magnetic resonance imaging (MRI) provides good soft-tissue contrast and no ionizing radiation but takes a longer scanning time, which leads to patient un- comfort and susceptibility to motion artifacts. To accelerate MRI scanning, under-sampling in the k-space (frequency space) is necessary but it violates the Nyquist sampling theorem and generates aliasing and blurring artifacts, making it challenging to reconstruct images. A previous study proposed a deep learning-based reconstruction model De-Aliasing Generative Adversarial Network (DAGAN) and showed competence over other reconstruction techniques. The goal of this project is to further optimize the DAGAN model to accomplish better reconstruction results.