De-noising Multi-coil Magnetic Resonance Imaging Using Patch-Based Adaptive Filtering in Wavelet Domain

Аннотация

Magnetic resonance imaging (MRI) frequently requires transform domain de-noising methods to preserve important features in the reconstructed images such as corners, sharp structures, and edges. Wavelet transform-based image de-noising is a standard approach used in MRI to recover smooth surface and sharp edges from the given noisy MR images, thereby improving diagnostic interpretations. Parallel magnetic resonance imaging (pMRI) techniques such as SENSE have been recently developed with an aim to improve the data acquisition speed, signal-to-noise ratio (SNR), and spatial resolution of the reconstructed images. However, the SENSE reconstruction algorithm often encounters noise during data acquisition and reconstruction process which not only contaminates the quality of the reconstructed images but also leads to poor diagnostic interpretations in clinical settings. During SENSE reconstruction process, noise can appear in the reconstructed image mainly due to two reasons (1) imperfections in the receiver coils; (2) un-folding the aliased images of multiple receiver coils to obtain a single composite image. In this paper, a new adaptive patch-based filtering in wavelet domain is presented to recover sharp structures and edges without introducing any artifacts in the SENSE reconstructed images. The proposed method uses soft-thresholding function as a shrinkage process which typically involves thresholding the small wavelet coefficients to reduce the noise without affecting the important features in the SENSE reconstructed images. For the evaluation of the proposed method, several experiments are performed using simulated phantom and in vivo data sets. The SENSE reconstruction quality using the proposed method is compared with contemporary de-nosing techniques, in terms of structural similarity index (SSIM) and peak signal-to-noise ratio (PSNR). Experimental results demonstrate that the SENSE reconstruction using the proposed method when compared to the other contemporary de-nosing methods successfully removes the noise and preserves the fine details in the reconstructed MR images without introducing blurring artifacts.