Improved image reconstruction from sensitivity-encoded data by wavelet denoising and Tokhonov regularization

Parallel magnetic resonance imaging through sensitivity encoding using multiple receiver coils has emerged as an effective tool to reduce imaging time. However, errors in both the estimated coil sensitivity maps and the measured data, and the ill-conditioned nature of the coefficient matrix (often associated with non-localized coils) can degrade image quality significantly, limiting speed enhancements. In this paper, we propose to use wavelet denoising to reduce noise in the coil sensitivity maps and a specially-designed Tikhonov regularization scheme to solve the ill-conditioned matrix equation. Experimental results show that these techniques produce significantly better images (with an improved signal-to-noise ratio and reduced aliasing artifacts) than conventional reconstruction methods based on matrix inversion with a diagonal regularization matrix.