An Adaptive SVD based De-Noising Filtering Scheme for parallel MRI

Magnetic Resonance Imaging (MRI) usually encounters noise during data acquisition and reconstruction process which not only contaminates the quality of reconstructed images but also leads to poor diagnostic interpretations in clinical setting. Parallel MRI have been evolved as an alternative technique in MRI to reduce the data acquisition time by acquiring a fewer number of data samples in k-space. However, the reconstruction quality of pMRI method such as generalized auto calibrating partially parallel acquisitions (GRAPPA) is still affected by spatially varying noise levels. During GRAPPA reconstruction process, noise can occur in the reconstructed image mainly due to the two reasons (i) imperfection in the receiver coils and (ii) due to combining the individual receiver coil images to obtain a single composite GRAPPA reconstructed image. Thus, there is an utter demand to de-noised the pMRI based reconstructed image by preserving the finer detail without degrading the quality of the reconstructed image. This paper proposes an innovative solution for de-noising the image by using new adaptive SVD based filtering method. This proposed method is applied in wavelet domain, on the GRAPPA reconstructed image to effectively reduce the noise while preserving the finer detail and edges information of the reconstructed image. The proposed method is evaluated on synthetic simulated brain data by performing GRAPPA reconstruction using acceleration factor of 2x and 4x. The reconstruction quality of the proposed method is evaluated and compared with other contemporary de-nosing techniques in terms of peak signal-to-noise ratio (PSNR) and RMSE in the reconstructed images. Experiment results demonstrates that the proposed method when compared to the other contemporary de-nosing methods, successfully removes the noise from the GRAPPA reconstructed images and preserves the fine details without inducing blurring artifacts in the reconstructed images.