Noise-robust reconstruction for accelerated MRI using contrastive learning

Abstract

Accelerated MRI method can reduce scan time by utilizing multichannel k-space data with sparse sampling. Simple zero-filling in the missing k-space data causes aliasing artifacts in the reconstructed image. Consequently, parallel imaging or compressed sensing methods were developed for image reconstruction that resolves aliasing artifacts. Recently, deep learning-based accelerated MRI reconstruction methods have shown outstanding performance in terms of computation time as well as image quality. However, conventional deep learning methods do not consider noise in the acquired data. The corruption due to noise may lead to wrong diagnosis in clinical practices. To address this issue, this thesis proposes a noise-robust reconstruction method using contrastive learning framework. At first, the encoder in the first stage is trained to extract representation features containing information about noise level in the input image. This is followed by the reconstruction network in the second stage, which takes noisy undersampled image and extracted features from the first stage as inputs to reconstruct alias-free image. Not only does the proposed method reconstruct higher quality images with well-preserved details than the baseline models, but also achieves superior quantitative results. The representation features extracted from the trained encoder contain content-invariant noise level information, thus can be applied to MR image reconstruction tasks of various datasets.