Abstract
Magnetic resonance imaging (MRI) is a useful technique for clinical diagnostics. However, MRI requires relatively long data acquisition, and the image quality can be deteriorated by physiological and thermal noises. Denoising is an important post-processing step for MRI, particularly when quantitative measurements are involved. Traditional denoising algorithms are mostly based on various spatial filtering. In recent years, CNN based algorithms have been explored for denoising of MRI magnitude images. In this study, we take one step forward by combining RDN with CBAM to construct a more robust framework for denoising of MR images. We trained the model with simulated MRI data with known “ground truth”. To illustrate the effectiveness of the model, we compared the performance of the model with other two state-of-the-art baseline frameworks.