Model-Based Iterative Reconstruction for Echo Planar Imaging: Methods and Applications

摘要

Echo planar imaging is widely-used clinical settings, for its speed. EPI acquisitions typically employ complex sampling protocols, and correspondingly require advanced reconstruction plans both to account for non-uniform data spacing and for managing system imperfections such as B0-field inhomogeneity, gradient non-linearity, and eddy currents, as well as subject-induced non-idealities like susceptibility variation. Such non-idealities are typically managed after the k-space to image domain transformation, which generates geometrical inaccuracies and blurring in images. In this work, we develop and investigate a comprehensive model-based iterative reconstruction (MBIR) framework that prospectively accounts for multiple non-idealities in accelerated single-shot EPI. When necessary, we also employ nonlinear regularization to mitigate noise amplification.