A model-based deep learning (DL) approach for denoising MR images

Motivation: The most prominent approach for MRI denoising has been the direct DL denoising methods. Robustness of such methods are not guaranteed while denoising MR images with unseen noise levels. Goal(s): Propose and evaluate a model-based DL denoising method for MR image. Approach: In this work, we adapt the model-based DL methods for MRI reconstruction to perform MR image denoising. Results: The proposed method was evaluated on in-vivo MRI data. The evaluation dataset consisted of MRI data of several contrasts for multiple anatomies. A qualitative analysis shows that the denoised images have fine structures well preserved. Impact: A model-based approach for MRI denoising provides guarantees against contrast and fine structure alterations during denoising. By virtue of the proposed method, we shall have robust and generalizable MRI denoisers.