Chemical species separation with simultaneous estimation of field map and T2* using a k-space formulation

Abstract

Chemical species separation techniques in image space are prone to incorporate several distortions. Some of these are signal accentuation in borders and geometrical warping from field inhomogeneity. These errors come from neglecting intraecho time variations. In this work, we present a new approach for chemical species separation in MRI with simultaneous estimation of field map and T2* decay, formulated entirely in k-space. In this approach, the time map is used to model the phase accrual from off-resonance precession and also the amplitude decay due to T2*. Our technique fits the signal model directly in k-space with the acquired data minimizing the l(2)-norm with an interior-point algorithm. Standard two dimensional gradient echo sequences in the thighs and head were used for demonstrating the technique. With this approach, we were able to obtain excellent estimation for the species, the field inhomogeneity, and T2* decay images. The results do not suffer from geometric distortions derived from the chemical shift or the field inhomogeneity. Importantly, as the T2* map is well positioned, the species signal in borders is correctly estimated. Considering intraecho time variations in a complete signal model in k-space for separating species yields superior estimation of the variables of interest when compared to existing methods.