• magnetic resonance imaging;
  • chemical species separation;
  • fat and water;
  • field inhomogeneities;
  • off–resonance correction;
  • Tmath image map


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 Tmath image 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 Tmath image. Our technique fits the signal model directly in k-space with the acquired data minimizing the l2-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 Tmath image decay images. The results do not suffer from geometric distortions derived from the chemical shift or the field inhomogeneity. Importantly, as the Tmath image 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. Magn Reson Med, 2012. © 2011 Wiley Periodicals, Inc.