Total Variation Regularized SENSE Image Reconstruction Based on Improved regularization parameter selection

In parallel magnetic resonance imaging (MRI), total variation (TV) regularized sensitivity coding (SENSE) reconstruction is a kind of fast and effective image reconstruction method. However, in the process of TV-SENSE reconstruction, the value of regularization parameter is an important factor that affects the signal-to-noise ratio (SNR) of image and directly affects the quality of image reconstruction. To solve the unbalanced problems of low signal-to-noise ratio and insufficient detail information preservation, this paper presents a regularization parameter selection method based on Monte-Carlo SURE in TV-SENSE image reconstruction. Adaptive iteration is adopted to select the regularization parameter value which minimizes the mean square error (MSE) of reconstructed image, so that achieve the balance between regularization term and data fidelity term. The reconstruction experiments were carried out by collecting the MRI phantom data under different acceleration factors, and the reconstruction results were compared with the generalized cross-validation (GCV) TV-SENSE method and L-curve TV-SENSE method. The research results show that the TV-SENSE image reconstruction algorithm based on the improved regularization parameter selection can effectively suppress the artifacts, the peak signal-to-noise ratio (PSNR) has also been improved and the artifact power (AP) of reconstructed image is also decreased even in the case of a higher acceleration factor, thus further verifies the feasibility and effectiveness of the proposed TV-SENSE image reconstruction algorithm.