Robust GRAPPA Reconstruction and Its Evaluation with Perceptual Difference Model (PDM)

Introduction GRAPPA [1] is a popular SMASH-type parallel imaging reconstruction technique. It allows one to extract the sensitivity information, or the fitting coefficients from the acquired ACS (Auto Calibration Signal) lines, and reconstruct the missing k-space lines from these coefficients. GRAPPA uses standard least-squares to get the “fitting” coefficients. We hypothesize that there are errors resulting from this estimation can be reduced using robust fitting techniques. We call this approach Robust GRAPPA. So as to easily test a variety of independent variables (image data set, noise, reduction factor, etc.) as a function of the type of robust estimation technique, we quantify image quality using a Perceptual Difference Model (PDM) [2-3] and evaluate 7,500 images.