Unsupervised MRI Images Denoising via Decoupled Expression

Abstract Magnetic Resonance Imaging (MRI) is widely adopted in medical diagnosis. Due to the spatial coding scheme, MRI image degraded by various noise. Recently, massive methods have been applied denoising. However, they lack consideration of artifacts images. In this paper, we propose an unsupervised denoising method called UEGAN based on decoupled expression. We decouple content and noise a noisy using encoders encoders. employ noising branch push decoder only extract The cycle-consistency loss ensures that denoised results match original To acquire visually realistic generations, add adversarial results. Image quality penalty helps retain rich details. perform experiments unpaired images from Brainweb datesets, achieve superior performances compared several popular approaches.