Joshua D. Trzasko1, Yunhong Shu2, Armando Manduca1, Matt A. Bernstein2
1Physiology and Biomedical Engineering, Mayo Clinic, Rochester, MN, United States; 2Department of Radiology, Mayo Clinic, Rochester, MN, United States
Least-norm reconstruction of undersampled Cartesian MRI data, often referred to a zero filling, remains a popular strategy due to its simple and efficient implementation, and readily-characterized behavior. For non-Cartesian MRI, however, determination of an analogous reconstruction is computationally demanding and requires iterative methods that may be impractical for clinical use. In this work, we propose a novel and efficient numerical framework based on positive semi-definite constrained least-squares regression for generating accurate, non-iterative (i.e., direct) approximators of least-norm reconstructions of non-Cartesian MRI data.